[go: up one dir, main page]
Next Issue
Volume 22, July-1
Previous Issue
Volume 22, June-1
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
7.3
3.4
Journals
Sensors
Volume 22
Issue 12
sensors-logo

Journal Menu

► Journal Menu

Journal Browser

► Journal Browser
share announcement

Sensors, Volume 22, Issue 12 (June-2 2022) – 346 articles

Cover Story (view full-size image): In this work, we suggested an innovative non-invasive microwave method for biosensing adherent cancer cells with different aggressiveness by measuring their dielectric properties. The sensor was designed according to its application with culture dishes and realized to test two groups of different cancer cell lines with different aggressiveness. Experimental results showed that the proposed sensor exhibited high sensitivity in the measurement of resonant frequency, which allowed discriminating between low- and high-metastatic cells, even from different types of cancer, paving the way to the development of more complex systems for non-invasive cancer tissue detection and characterization. View this paper
  • Issues are regarded as officially published after their release is announced to the table of contents alert mailing list.
  • You may sign up for e-mail alerts to receive table of contents of newly released issues.
  • PDF is the official format for papers published in both, html and pdf forms. To view the papers in pdf format, click on the "PDF Full-text" link, and use the free Adobe Reader to open them.
Order results
Result details
Section
Show export options expand_more Show export options expand_less
Select all
Export citation of selected articles as:
18 pages, 6144 KiB  
Article
An Embedded Portable Lightweight Platform for Real-Time Early Smoke Detection
by Bowen Liu, Bingjian Sun, Pengle Cheng and Ying Huang
Sensors 2022, 22(12), 4655; https://doi.org/10.3390/s22124655 - 20 Jun 2022
Cited by 6 | Viewed by 3135
Abstract
The advances in developing more accurate and fast smoke detection algorithms increase the need for computation in smoke detection, which demands the involvement of personal computers or workstations. Better detection results require a more complex network structure of the smoke detection algorithms and [...] Read more.
The advances in developing more accurate and fast smoke detection algorithms increase the need for computation in smoke detection, which demands the involvement of personal computers or workstations. Better detection results require a more complex network structure of the smoke detection algorithms and higher hardware configuration, which disqualify them as lightweight portable smoke detection for high detection efficiency. To solve this challenge, this paper designs a lightweight portable remote smoke front-end perception platform based on the Raspberry Pi under Linux operating system. The platform has four modules including a source video input module, a target detection module, a display module, and an alarm module. The training images from the public data sets will be used to train a cascade classifier characterized by Local Binary Pattern (LBP) using the Adaboost algorithm in OpenCV. Then the classifier will be used to detect the smoke target in the following video stream and the detected results will be dynamically displayed in the display module in real-time. If smoke is detected, warning messages will be sent to users by the alarm module in the platform for real-time monitoring and warning on the scene. Case studies showed that the developed system platform has strong robustness under the test datasets with high detection accuracy. As the designed platform is portable without the involvement of a personal computer and can efficiently detect smoke in real-time, it provides a potential affordable lightweight smoke detection option for forest fire monitoring in practice. Full article
(This article belongs to the Topic Applications in Image Analysis and Pattern Recognition)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>System overall design block diagram.</p> Full article ">Figure 2
<p>Raspberry Pi 4B Microprocessor.</p> Full article ">Figure 3
<p>Hardware Design Block Diagram.</p> Full article ">Figure 4
<p>Example of filtering effect: (<b>a</b>,<b>b</b>): original image; (<b>c</b>,<b>d</b>): mean filtering; (<b>e</b>,<b>f</b>): median filtering; (<b>g</b>,<b>h</b>): Gaussian filtering.</p> Full article ">Figure 5
<p>Histogram equalization of pixels.</p> Full article ">Figure 6
<p>Histogram equalization: (<b>a</b>) original histogram of RGB channel, equalized histogram; (<b>b</b>) original image of RGB channel, equalized image; (<b>c</b>) original image, equalized image.</p> Full article ">Figure 7
<p>Original LBP operator.</p> Full article ">Figure 8
<p>Cascade classifier structure diagram.</p> Full article ">Figure 9
<p>Email alarm interface.</p> Full article ">Figure 10
<p>Remote connection interface.</p> Full article ">Figure 11
<p>Positive sample examples.</p> Full article ">Figure 12
<p>Negative sample examples.</p> Full article ">Figure 13
<p>Example of detection effect where regions in green boxes are detected as smoke region: (<b>a</b>) 307th frame of test video 1 (<b>b</b>) 674th frame of test video 1 (<b>c</b>) 237th frame of test video 2 (<b>d</b>) 778th frame of test video 2 (<b>e</b>) 807th frame of test video 3 (<b>f</b>) 1414th frame of test video 3 (<b>g</b>) 454th frame of test video 7 (<b>h</b>) 584th frame of test video 7 (<b>i</b>) 698th frame of test video 8 (<b>j</b>) 770th frame of test video 8.</p> Full article ">
26 pages, 12941 KiB  
Article
Conceptual Modeling of Extended Collision Warning System from the Perspective of Smart Product-Service System
by Chunlong Wu, Hanyu Lv, Tianming Zhu, Yunhe Liu and Marcus Vinicius Pereira Pessôa
Sensors 2022, 22(12), 4654; https://doi.org/10.3390/s22124654 - 20 Jun 2022
Cited by 5 | Viewed by 2268
Abstract
While Product-Service Systems (PSS) have a potential sustainability impact by increasing a product’s life and reducing resource consumption, the lack of ownership might lead to less responsible user behavior. Smart PSS can overcome this obstacle and guarantee correct and safe PSS use. In [...] Read more.
While Product-Service Systems (PSS) have a potential sustainability impact by increasing a product’s life and reducing resource consumption, the lack of ownership might lead to less responsible user behavior. Smart PSS can overcome this obstacle and guarantee correct and safe PSS use. In this context, intelligent connected vehicles (ICVs) with PSS can effectively reduce traffic accidents and ensure the safety of vehicles and pedestrians by guaranteeing optimal and safe vehicle operation. A core subsystem to support that is the collision-warning system (CWS). Existing CWSs are, however, limited to in-car warning; users have less access to the warning information, so the result of CWS for collision avoidance is insufficient. Therefore, CWS needs to be extended to include more elements and stakeholders in the collision scenario. This paper aims to provide a novel understanding of extended CWS (ECWS), outline the conceptual framework of ECWS, and contribute a conceptual modeling approach of ECWS from the smart PSS perspective at the functional level. It defines an integrated solution of intelligent products and warning services. The function is modeled based on the Theory of Inventive Problem Solving (TRIZ). Functions of an ECWS from the perspective of smart PSS can be comprehensively expressed to form an overall solution of integrated intelligent products, electronic services, and stakeholders. Based on the case illustration, the proposed method can effectively help function modeling and development of the ECWS at a conceptual level. This can effectively avoid delays due to traffic accidents and ensure the safety of vehicles and pedestrians. Full article
(This article belongs to the Special Issue Sensor Fusion and Advanced Controller for Connected and Automated Vehicles)
►▼ Show Figures

Figure 1

Figure 1
<p>Stakeholders–Intelligent Product–Intelligent Service system.</p> Full article ">Figure 2
<p>Conceptual framework of extended CWS.</p> Full article ">Figure 3
<p>Elements of the TRIZ function model—fill contents in the chart.</p> Full article ">Figure 4
<p>Proposed modeling process of the extended CWS.</p> Full article ">Figure 5
<p>Functional decomposition process.</p> Full article ">Figure 6
<p>Cloud platform structure.</p> Full article ">Figure 7
<p>Overall architecture of extended CWS from the perspective of smart PSS.</p> Full article ">Figure 8
<p>(<b>a</b>) Dongfeng S50 EV. (<b>b</b>) Sightseeing vehicle.</p> Full article ">Figure 9
<p>Experimental Area.</p> Full article ">Figure 10
<p>Extended CWS from the perspective of smart PSS.</p> Full article ">Figure 11
<p>Functions of extended CWS from the perspective of smart PSS.</p> Full article ">Figure 12
<p>Function model of the intelligent product system.</p> Full article ">Figure 13
<p>Function model of stakeholders.</p> Full article ">Figure 14
<p>Function model of vehicle collision-warning service system.</p> Full article ">Figure 15
<p>Function model of vehicle–pedestrian collision-warning service system.</p> Full article ">Figure 16
<p>The cloud platform of the extended CWS.</p> Full article ">Figure 17
<p>The function model of the extended CWS.</p> Full article ">Figure 18
<p>Parts of the indoor bench installation of the system.</p> Full article ">Figure 19
<p>Monitoring platform.</p> Full article ">Figure 20
<p>Visualization interface based on CANape17.0.</p> Full article ">Figure 21
<p><span class="html-italic">Relative_Dist_Forward</span> of two vehicles.</p> Full article ">Figure 22
<p><span class="html-italic">TTC_Forward</span> of two vehicles.</p> Full article ">Figure 23
<p>Traffic lights recognition based on V2X.</p> Full article ">Figure 24
<p>Road information publishing based on V2X.</p> Full article ">Figure 25
<p>Pedestrian recognition and avoidance based on V2X.</p> Full article ">Figure 26
<p>Intersection collision warning based on V2X.</p> Full article ">Figure 27
<p>Forward collision warning based on V2X.</p> Full article ">
17 pages, 3259 KiB  
Review
Recent Advances in Flexible Sensors and Their Applications
by Bouchaib Zazoum, Khalid Mujasam Batoo and Muhammad Azhar Ali Khan
Sensors 2022, 22(12), 4653; https://doi.org/10.3390/s22124653 - 20 Jun 2022
Cited by 79 | Viewed by 19282
Abstract
Flexible sensors are low cost, wearable, and lightweight, as well as having a simple structure as per the requirements of engineering applications. Furthermore, for many potential applications, such as human health monitoring, robotics, wearable electronics, and artificial intelligence, flexible sensors require high sensitivity [...] Read more.
Flexible sensors are low cost, wearable, and lightweight, as well as having a simple structure as per the requirements of engineering applications. Furthermore, for many potential applications, such as human health monitoring, robotics, wearable electronics, and artificial intelligence, flexible sensors require high sensitivity and stretchability. Herein, this paper systematically summarizes the latest progress in the development of flexible sensors. The review briefly presents the state of the art in flexible sensors, including the materials involved, sensing mechanisms, manufacturing methods, and the latest development of flexible sensors in health monitoring and soft robotic applications. Moreover, this paper provides perspectives on the challenges in this field and the prospect of flexible sensors. Full article
(This article belongs to the Special Issue Wearable Sensors for Physical Activity Monitoring and Motion Control)
►▼ Show Figures

Figure 1

Figure 1
<p>The main parts of a flexible sensor comprise flexible materials and applications [<a href="#B16-sensors-22-04653" class="html-bibr">16</a>]. Copyright 2017, Elsevier.</p> Full article ">Figure 2
<p>Application of the AuNWs/latex strain sensor in human motion monitoring: (<b>a</b>) movement of the forearm muscles; (<b>b</b>) movement of the cheeks; (<b>c</b>) continuous throat movement while saying “hello”; and (<b>d</b>) detecting human wrist pulses [<a href="#B66-sensors-22-04653" class="html-bibr">66</a>]. Copyright 2015, John Wiley and Sons.</p> Full article ">Figure 3
<p>(<b>a</b>) An analysis of the capacitance at 0–140 kPa as a function of applied pressure for crack-enhanced microfluidic pressure sensors prepared with five sensing liquids. (<b>b</b>) Normalized capacitance plotted as a function of all sensors. The values simulated are represented by solid lines. (<b>c</b>) Normalized capacitance of EG or water-based sensors as a function of train, ranging from 0% to 9%. (<b>d</b>) The channel and crack simulated in the COMSOL are depicted schematically in this diagram. (<b>e</b>) In each liquid example, COMSOL modeling results were achieved under the initial conditions, simulating the interface in the crack after filling the channel. (<b>f</b>) Angles of contact between liquids and the PDMS surface. (<b>g</b>) COMSOL modeling results produced by COMSOL for the water and PC under various pressure settings. (<b>h</b>) Simulated Δd/d<sub>0</sub> vs. the applied pressure corresponding to each. (<b>i</b>) Wettability parameters (k) plotted as function of contact angle. (<b>j</b>) Illustration of the simulated normalized capacitance as function of the pressure [<a href="#B83-sensors-22-04653" class="html-bibr">83</a>]. Copyright 2017, American Chemical Society.</p> Full article ">Figure 4
<p>Sketch of main components of a triboelectric nano-generator, including the charge-generating layer, the charge-trapping layer, the charge-collecting layer, as well as the charge-storage layer. Reproduced with permission [<a href="#B109-sensors-22-04653" class="html-bibr">109</a>]. Copyright 2020, Springer Nature.</p> Full article ">Figure 5
<p>Human body biosignals and the healthcare sensors that measure them. Reproduced with permission [<a href="#B120-sensors-22-04653" class="html-bibr">120</a>]. Copyright 2018, Royal Society of Chemistry.</p> Full article ">Figure 6
<p>Radial artery pulse waves and characterizations of the capacitive pressure response. (<b>a</b>–<b>d</b>) The radial artery pulse waves were monitored using four different types of devices with varied geometry. (<b>e</b>) Statistical data on the change in capacitance caused by applying varying amounts of shear pressures to different sensors. (<b>f</b>) Curves of relaxation and steady state after loading and unloading different types of sensors [<a href="#B121-sensors-22-04653" class="html-bibr">121</a>]. Copyright 2014, John Wiley and Sons.</p> Full article ">
13 pages, 4709 KiB  
Communication
High-Resolution Hyperspectral Imaging Using Low-Cost Components: Application within Environmental Monitoring Scenarios
by Mary B. Stuart, Matthew Davies, Matthew J. Hobbs, Tom D. Pering, Andrew J. S. McGonigle and Jon R. Willmott
Sensors 2022, 22(12), 4652; https://doi.org/10.3390/s22124652 - 20 Jun 2022
Cited by 43 | Viewed by 8671
Abstract
High-resolution hyperspectral imaging is becoming indispensable, enabling the precise detection of spectral variations across complex, spatially intricate targets. However, despite these significant benefits, currently available high-resolution set-ups are typically prohibitively expensive, significantly limiting their user base and accessibility. These limitations can have wider [...] Read more.
High-resolution hyperspectral imaging is becoming indispensable, enabling the precise detection of spectral variations across complex, spatially intricate targets. However, despite these significant benefits, currently available high-resolution set-ups are typically prohibitively expensive, significantly limiting their user base and accessibility. These limitations can have wider implications, limiting data collection opportunities, and therefore our knowledge, across a wide range of environments. In this article we introduce a low-cost alternative to the currently available instrumentation. This instrument provides hyperspectral datasets capable of resolving spectral variations in mm-scale targets, that cannot typically be resolved with many existing low-cost hyperspectral imaging alternatives. Instrument metrology is provided, and its efficacy is demonstrated within a mineralogy-based environmental monitoring application highlighting it as a valuable addition to the field of low-cost hyperspectral imaging. Full article
(This article belongs to the Special Issue Hyperspectral Imaging Sensing and Analysis)
►▼ Show Figures

Figure 1

Figure 1
<p>Schematic diagram of the Low-Cost High-Resolution hyperspectral imager showing how axial and marginal rays pass through the optical system. Blue, green, and red lines represent example wavelength rays after diffraction has taken place. Not to scale.</p> Full article ">Figure 2
<p>Example frames of an ammonite fossil taken from a hyperspectral data cube demonstrating the spatial resolution possible with this instrument. The first panel shows a standard color image of the target for reference. The additional panels show hyperspectral frames captured at focal lengths of 18 mm and 55 mm, respectively.</p> Full article ">Figure 3
<p>The Low-Cost High-Resolution hyperspectral imager within a laboratory setting.</p> Full article ">Figure 4
<p>Workflow used to capture a hyperspectral image with the Low-Cost High-Resolution instrument detailing image acquisition and post processing stages.</p> Full article ">Figure 5
<p>Spectrum captured from a Mercury Argon lamp using the Low-Cost High-Resolution instrument highlighting the peaks present at 546.074 nm and 576.960 nm that were used to spectrally calibrate the instrument.</p> Full article ">Figure 6
<p>CTF analysis for both focal lengths. (<b>A</b>,<b>B</b>) (<b>left</b>) show an image frame of the resolution target captured at an 18 mm focal length and a 55 mm focal length, respectively, (<b>C</b>) (<b>right</b>) shows the resulting CTF values for horizontal line pairs.</p> Full article ">Figure 7
<p>Knife-edge measurements for each focal length. (<b>A</b>) shows the results for the 18 mm focal length demonstrating a one-pixel discrepancy between orientations. (<b>B</b>) shows results for the 55 mm focal length demonstrating a two-pixel discrepancy between orientations.</p> Full article ">Figure 8
<p>Hyperspectral image frames of a gneiss sample demonstrating the spatial resolution of this instrument. Characteristic banding and surface features are clearly visible within the hyperspectral data and can be easily related to their specific location on the original target. The image on the left is a standard color image of the sample and the hyperspectral images are on the right-hand side of the figure. The hyperspectral images are just one slice through the data cube that contains 689 discrete wavelength values. RGB frames represent the availability of different wavelength frames within the hyperspectral data cube.</p> Full article ">Figure 9
<p>Two hyperspectral image frames of a basalt sample compared to standard color images. Note the clarity of the surface features within the hyperspectral frames allowing clear differentiation between feldspar and surface features. The hyperspectral images are just one slice through the data cube that contains 689 discrete wavelength values. RGB frames represent the availability of different wavelength frames within the hyperspectral data cube.</p> Full article ">Figure 10
<p>Spectral data for a piece of supraglacial debris with orange pigmentation. (<b>A</b>) shows a standard color image of the rock sample highlighting the approximate locations that correspond to the spectral curves shown in (<b>B</b>).</p> Full article ">Figure 11
<p>Spectral and spatial information obtained for a sample of lapis lazuli. Note the expected increase in reflectance across blue wavelengths followed by a steady reduction in reflectance towards longer wavelengths. The hyperspectral images represent single slices through the data cube that contains 689 discrete wavelength values. The reconstructed RGB image is created using red-green-blue equivalent images taken from the hyperspectral data cube.</p> Full article ">Figure 12
<p>Spectral data obtained from a sample of lapis lazuli. Deviations from the laboratory-measured spectrum are associated with regions of low signal within the illumination spectrum. Note the correlation between the spectral response curve and the spectral-spatial data shown in <a href="#sensors-22-04652-f011" class="html-fig">Figure 11</a>.</p> Full article ">
19 pages, 3372 KiB  
Article
Estimation of Knee Extension Force Using Mechanomyography Signals Based on GRA and ICS-SVR
by Zebin Li, Lifu Gao, Wei Lu, Daqing Wang, Huibin Cao and Gang Zhang
Sensors 2022, 22(12), 4651; https://doi.org/10.3390/s22124651 - 20 Jun 2022
Cited by 6 | Viewed by 2489
Abstract
During lower-extremity rehabilitation training, muscle activity status needs to be monitored in real time to adjust the assisted force appropriately, but it is a challenging task to obtain muscle force noninvasively. Mechanomyography (MMG) signals offer unparalleled advantages over sEMG, reflecting the intention of [...] Read more.
During lower-extremity rehabilitation training, muscle activity status needs to be monitored in real time to adjust the assisted force appropriately, but it is a challenging task to obtain muscle force noninvasively. Mechanomyography (MMG) signals offer unparalleled advantages over sEMG, reflecting the intention of human movement while being noninvasive. Therefore, in this paper, based on MMG, a combined scheme of gray relational analysis (GRA) and support vector regression optimized by an improved cuckoo search algorithm (ICS-SVR) is proposed to estimate the knee joint extension force. Firstly, the features reflecting muscle activity comprehensively, such as time-domain features, frequency-domain features, time–frequency-domain features, and nonlinear dynamics features, were extracted from MMG signals, and the relational degree was calculated using the GRA method to obtain the correlation features with high relatedness to the knee joint extension force sequence. Then, a combination of correlated features with high relational degree was input into the designed ICS-SVR model for muscle force estimation. The experimental results show that the evaluation indices of the knee joint extension force estimation obtained by the combined scheme of GRA and ICS-SVR were superior to other regression models and could estimate the muscle force with higher estimation accuracy. It is further demonstrated that the proposed scheme can meet the need of muscle force estimation required for rehabilitation devices, powered prostheses, etc. Full article
(This article belongs to the Topic Human–Machine Interaction)
►▼ Show Figures

Figure 1

Figure 1
<p>The block diagram of knee joint extension force estimation.</p> Full article ">Figure 2
<p>The procedure of ICS-SVR.</p> Full article ">Figure 3
<p>The search space of the four test benchmark functions.</p> Full article ">Figure 4
<p>Comparison of convergence curves of the four algorithms for the four test benchmark functions.</p> Full article ">Figure 5
<p>GRA analysis of MMG features of subject S1.</p> Full article ">Figure 6
<p>Results of knee joint extension force estimation for subject S1 with feature combination D.</p> Full article ">Figure 7
<p>Statistical analysis of knee joint extension force estimation with different regression models. (<b>a</b>) The <span class="html-italic">R</span>-value of the force estimation against the actual observed values. (<b>b</b>) The RMSE of the force estimation against the actual observed values. (<b>c</b>) The MAPE of the force estimation against the actual observed values.</p> Full article ">
13 pages, 2858 KiB  
Article
Real-Time Sound Source Localization for Low-Power IoT Devices Based on Multi-Stream CNN
by Jungbeom Ko, Hyunchul Kim and Jungsuk Kim
Sensors 2022, 22(12), 4650; https://doi.org/10.3390/s22124650 - 20 Jun 2022
Cited by 9 | Viewed by 3732
Abstract
Voice-activated artificial intelligence (AI) technology has advanced rapidly and is being adopted in various devices such as smart speakers and display products, which enable users to multitask without touching the devices. However, most devices equipped with cameras and displays lack mobility; therefore, users [...] Read more.
Voice-activated artificial intelligence (AI) technology has advanced rapidly and is being adopted in various devices such as smart speakers and display products, which enable users to multitask without touching the devices. However, most devices equipped with cameras and displays lack mobility; therefore, users cannot avoid touching them for face-to-face interactions, which contradicts the voice-activated AI philosophy. In this paper, we propose a deep neural network-based real-time sound source localization (SSL) model for low-power internet of things (IoT) devices based on microphone arrays and present a prototype implemented on actual IoT devices. The proposed SSL model delivers multi-channel acoustic data to parallel convolutional neural network layers in the form of multiple streams to capture the unique delay patterns for the low-, mid-, and high-frequency ranges, and estimates the fine and coarse location of voices. The model adapted in this study achieved an accuracy of 91.41% on fine location estimation and a direction of arrival error of 7.43° on noisy data. It achieved a processing time of 7.811 ms per 40 ms samples on the Raspberry Pi 4B. The proposed model can be applied to a camera-based humanoid robot that mimics the manner in which humans react to trigger voices in crowded environments. Full article
(This article belongs to the Special Issue The Convergence of Artificial Intelligence, Sensor and System for Sustainable E-healthcare Services and Biomedical Applications)
►▼ Show Figures

Figure 1

Figure 1
<p>(<b>a</b>) Limitation of the camera vision-based method is that it cannot track if the user deviates from the field of view of the camera; (<b>b</b>) The advantage of the microphone array-based sound localization method is that it can track the user by estimating the direction through sound regardless of any direction.</p> Full article ">Figure 2
<p>System overview of real-time SSL system using IoT devices.</p> Full article ">Figure 3
<p>(<b>a</b>) A 2D schematic map of the microphone array and sound sources. The blue dots indicate sound sources, and the colored lines are extension lines between the center of the microphone array and each microphone; (<b>b</b>) Regions were divided according to the locations of the microphones.</p> Full article ">Figure 4
<p>(<b>a</b>) Block diagram of the model architecture, (<b>b</b>) conventional MS block, (<b>c</b>) efficient MS block, and (<b>d</b>) aggregation gate. The colored arrows indicate the stride of convolution, the subsequent batch normalization (BN) [<a href="#B17-sensors-22-04650" class="html-bibr">17</a>], and ReLU [<a href="#B18-sensors-22-04650" class="html-bibr">18</a>].</p> Full article ">Figure 5
<p>Comparison between (<b>a</b>) standard 1D Convolution and (<b>b</b>) depth-wise separable 1D convolution.</p> Full article ">Figure 6
<p>Comparison between (<b>a</b>) ACC and (<b>b</b>) DOA error in test data according to the type of block composing the SSL model. Here, “O” indicates efficient MS blocks, and “X” indicates conventional MS blocks.</p> Full article ">Figure 7
<p>Comparison between the inference times using a desktop CPU according to the type of block composing the SSL model and the framework used for inference.</p> Full article ">
25 pages, 1829 KiB  
Article
Degradation Detection in a Redundant Sensor Architecture
by Amer Kajmakovic, Konrad Diwold, Kay Römer, Jesus Pestana and Nermin Kajtazovic
Sensors 2022, 22(12), 4649; https://doi.org/10.3390/s22124649 - 20 Jun 2022
Cited by 6 | Viewed by 3753
Abstract
Safety-critical automation often requires redundancy to enable reliable system operation. In the context of integrating sensors into such systems, the one-out-of-two (1oo2) sensor architecture is one of the common used methods used to ensure the reliability and traceability of sensor readings. In taking [...] Read more.
Safety-critical automation often requires redundancy to enable reliable system operation. In the context of integrating sensors into such systems, the one-out-of-two (1oo2) sensor architecture is one of the common used methods used to ensure the reliability and traceability of sensor readings. In taking such an approach, readings from two redundant sensors are continuously checked and compared. As soon as the discrepancy between two redundant lines deviates by a certain threshold, the 1oo2 voter (comparator) assumes that there is a fault in the system and immediately activates the safe state. In this work, we propose a novel fault prognosis algorithm based on the discrepancy signal. We analyzed the discrepancy changes in the 1oo2 sensor configuration caused by degradation processes. Several publicly available databases were checked, and the discrepancy between redundant sensors was analyzed. An initial analysis showed that the discrepancy between sensor values changes (increases or decreases) over time. To detect an increase or decrease in discrepancy data, two trend detection methods are suggested, and the evaluation of their performance is presented. Moreover, several models were trained on the discrepancy data. The models were then compared to determine which of the models can be best used to describe the dynamics of the discrepancy changes. In addition, the best-fitting models were used to predict the future behavior of the discrepancy and to detect if, and when, the discrepancy in sensor readings will reach a critical point. Based on the prediction of the failure date, the customer can schedule the maintenance system accordingly and prevent its entry into the safe state—or being shut down. Full article
(This article belongs to the Collection Recent Advances in Fault Diagnostics, Prognostics, and Intelligent Condition-Based Maintenance)
►▼ Show Figures

Figure 1

Figure 1
<p>Drift and delay or sensor response.</p> Full article ">Figure 2
<p>Model of the 1oo2D (1-out-of-2 with diagnostics) safety architecture.</p> Full article ">Figure 3
<p>Data generated from the publicly available dataset [<a href="#B14-sensors-22-04649" class="html-bibr">14</a>]: (<b>a</b>) Sensor measurements averaged per day for two sensors; (<b>b</b>) discrepancy signal.</p> Full article ">Figure 4
<p>Flowchart outlining the steps in the proposed approach.</p> Full article ">Figure 5
<p>Explanatory figure of the prediction error calculations.</p> Full article ">Figure 6
<p>Discrepancy of (<b>a</b>) redundant MOX gas sensors presented in [<a href="#B50-sensors-22-04649" class="html-bibr">50</a>], (<b>b</b>) redundant gas sensors [<a href="#B51-sensors-22-04649" class="html-bibr">51</a>], and (<b>c</b>) redundant chemical sensors presented in [<a href="#B46-sensors-22-04649" class="html-bibr">46</a>].</p> Full article ">Figure 7
<p>Discrepancy data for six pH sensors presented in [<a href="#B14-sensors-22-04649" class="html-bibr">14</a>], where graph (<b>a</b>) shows discrepancies of the identical sensors, and (<b>b</b>,<b>c</b>) show discrepancies among the rest of the sensors.</p> Full article ">Figure 8
<p>Discrepancy among the sensors’ readings, also showing calibration events on days 183, 365, and 541.</p> Full article ">Figure 9
<p>Boxplots of the root mean square error (RMSE) calculated for the trained models and categorized into four data periods.</p> Full article ">Figure 10
<p>Iterative process of trend detection with raw dataset. The graphs show (<b>a</b>) raw data, (<b>b</b>) trend detection, and (<b>c</b>) slope calculated by trend detection methods.</p> Full article ">Figure 11
<p>Recursive process of trend detection with filtered (Hampel filter) and averaged (window = ten) data. The graphs show (<b>a</b>) processed data (<b>b</b>), trend detection, and (<b>c</b>) the slope calculated when applying trend detection methods.</p> Full article ">Figure 12
<p>Example of the algorithm’s calculations for data record “T1aT2a”: (<b>a</b>) discrepancy data with threshold, (<b>b</b>) RMSE values for holdout values, (<b>c</b>) results of the voting, (<b>d</b>) prediction errors of both methods, and (<b>e</b>) final results of the algorithm based on the voting.</p> Full article ">Figure 13
<p>Prediction error for the last 60, 40, 20, and 10 days before threshold is reached, where green triangles represent mean values, for (<b>a</b>) period I ([0, 182] days), (<b>b</b>) period II ([183, 356] days), (<b>c</b>) period III ([356, 541] days), and (<b>d</b>) period IV ([542, 730] days).</p> Full article ">Figure 14
<p>Confusion matrices for event detection, where the presented approach is applied to (<b>a</b>) all days of the dataset ([0, 730] days); (<b>b</b>) days of the warranty period of the sensors ([0, 365] days).</p> Full article ">
17 pages, 30663 KiB  
Article
Algorithms and Methods for the Fault-Tolerant Design of an Automated Guided Vehicle
by Ralf Stetter
Sensors 2022, 22(12), 4648; https://doi.org/10.3390/s22124648 - 20 Jun 2022
Cited by 5 | Viewed by 3047
Abstract
Researchers around the globe have contributed for many years to the research field of fault-tolerant control; the importance of this field is ever increasing as a consequence of the rising complexity of technical systems, the enlarging importance of electronics and software as well [...] Read more.
Researchers around the globe have contributed for many years to the research field of fault-tolerant control; the importance of this field is ever increasing as a consequence of the rising complexity of technical systems, the enlarging importance of electronics and software as well as the widening share of interconnected and cloud solutions. This field was supplemented in recent years by fault-tolerant design. Two main goals of fault-tolerant design can be distinguished. The first main goal is the improvement of the controllability and diagnosability of technical systems through intelligent design. The second goal is the enhancement of the fault-tolerance of technical systems by means of inherently fault-tolerant design characteristics. Inherently fault-tolerant design characteristics are, for instance, redundancy or over-actuation. This paper describes algorithms, methods and tools of fault-tolerant design and an application of the concept to an automated guided vehicle (AGV). This application took place on different levels ranging from conscious requirements management to redundant elements, which were consciously chosen, on the most concrete level of a technical system, i.e., the product geometry. The main scientific contribution of the paper is a methodical framework for fault-tolerant design, as well as certain algorithms and methods within this framework. The underlying motivation is to support engineers in design and control trough product development process transparency and appropriate algorithms and methods. Full article
(This article belongs to the Special Issue Sensors and Fault-Tolerant Systems for Automated Guided Vehicles)
►▼ Show Figures

Figure 1

Figure 1
<p>Automated guided vehicle (AGV) and operation scenario.</p> Full article ">Figure 2
<p>Levels for characteristics of fault-tolerant design according to the levels of product concretization depicted in a V-model.</p> Full article ">Figure 3
<p>Functions of an AGV modelled in the IFM framework.</p> Full article ">Figure 4
<p>Accommodation of the fault “steering angle disturbance”.</p> Full article ">Figure 5
<p>Drive module of the AGV—torque steering and certain components.</p> Full article ">Figure 6
<p>Exemplary estimation result; fault at k = 6000 s.</p> Full article ">Figure 7
<p>Residuals generated by the virtual sensor.</p> Full article ">Figure 8
<p>Compensation factor generated with the fuzzy virtual actuator.</p> Full article ">Figure 9
<p>Example for physical phenomena for the realization of a function of a watch.</p> Full article ">Figure 10
<p>Physical effect chain for the steering system of the AGV.</p> Full article ">Figure 11
<p>Components of the drive module of the AGV.</p> Full article ">Figure 12
<p>Redundant ultrasonic sensors at the side of the AGV.</p> Full article ">
16 pages, 4513 KiB  
Article
Video Anomaly Detection Based on Convolutional Recurrent AutoEncoder
by Bokun Wang and Caiqian Yang
Sensors 2022, 22(12), 4647; https://doi.org/10.3390/s22124647 - 20 Jun 2022
Cited by 19 | Viewed by 3967
Abstract
As an essential task in computer vision, video anomaly detection technology is used in video surveillance, scene understanding, road traffic analysis and other fields. However, the definition of anomaly, scene change and complex background present great challenges for video anomaly detection tasks. The [...] Read more.
As an essential task in computer vision, video anomaly detection technology is used in video surveillance, scene understanding, road traffic analysis and other fields. However, the definition of anomaly, scene change and complex background present great challenges for video anomaly detection tasks. The insight that motivates this study is that the reconstruction error for normal samples would be lower since they are closer to the training data, while the anomalies could not be reconstructed well. In this paper, we proposed a Convolutional Recurrent AutoEncoder (CR-AE), which combines an attention-based Convolutional Long–Short-Term Memory (ConvLSTM) network and a Convolutional AutoEncoder. The ConvLSTM network and the Convolutional AutoEncoder could capture the irregularity of the temporal pattern and spatial irregularity, respectively. The attention mechanism was used to obtain the current output characteristics from the hidden state of each Covn-LSTM layer. Then, a convolutional decoder was utilized to reconstruct the input video clip and the testing video clip with higher reconstruction error, which were further judged to be anomalies. The proposed method was tested on two popular benchmarks (UCSD ped2 Dataset and Avenue Dataset), and the experimental results demonstrated that CR-AE achieved 95.6% and 73.1% frame-level AUC on two public datasets, respectively. Full article
(This article belongs to the Section Optical Sensors)
►▼ Show Figures

Figure 1

Figure 1
<p>Examples of the hand-crafted feature. (<b>a</b>) Object trajectory [<a href="#B24-sensors-22-04647" class="html-bibr">24</a>]. (<b>b</b>) Dense trajectory [<a href="#B26-sensors-22-04647" class="html-bibr">26</a>]. (<b>c</b>) Histograms of gradients (HOG) [<a href="#B31-sensors-22-04647" class="html-bibr">31</a>]. (<b>d</b>) Spatio-temporal video volumes (STVs) [<a href="#B30-sensors-22-04647" class="html-bibr">30</a>].</p> Full article ">Figure 2
<p>Examples of the deep learning-based method. (<b>a</b>) GMFC-VAE [<a href="#B32-sensors-22-04647" class="html-bibr">32</a>]. (<b>b</b>) GAN [<a href="#B18-sensors-22-04647" class="html-bibr">18</a>,<a href="#B33-sensors-22-04647" class="html-bibr">33</a>].</p> Full article ">Figure 3
<p>Overview of our proposed method.</p> Full article ">Figure 4
<p>Overall architecture of the proposed CR-AE model.</p> Full article ">Figure 5
<p>This is a figure. Schemes follow the same formatting.</p> Full article ">Figure 6
<p>ROC curves for the UCSD Ped2 dataset.</p> Full article ">Figure 7
<p>Visualization of the testing results.</p> Full article ">Figure 8
<p>Examples of better and worse abnormality detection results. (<b>a</b>) cars on the sidewalk. (<b>b</b>) cyclists on the sidewalk. (<b>c</b>) intense movements. (<b>d</b>) cars on the sidewalk. (<b>e</b>) occluded, cyclists on the sidewalk (<b>f</b>) poorly illuminated, cyclists on the sidewalk. (<b>g</b>) occluded, scooters on the sidewalk. (<b>h</b>) lost package.</p> Full article ">Figure 8 Cont.
<p>Examples of better and worse abnormality detection results. (<b>a</b>) cars on the sidewalk. (<b>b</b>) cyclists on the sidewalk. (<b>c</b>) intense movements. (<b>d</b>) cars on the sidewalk. (<b>e</b>) occluded, cyclists on the sidewalk (<b>f</b>) poorly illuminated, cyclists on the sidewalk. (<b>g</b>) occluded, scooters on the sidewalk. (<b>h</b>) lost package.</p> Full article ">
20 pages, 15513 KiB  
Article
A Single-Phase High-Impedance Ground Faulty Feeder Detection Method for Small Resistance to Ground Systems Based on Current-Voltage Phase Difference
by Zequan Hou, Zhihua Zhang, Yizhao Wang, Jiandong Duan, Wanying Yan and Wenchao Lu
Sensors 2022, 22(12), 4646; https://doi.org/10.3390/s22124646 - 20 Jun 2022
Cited by 6 | Viewed by 2058
Abstract
At present, the small resistance to ground system (SRGS) is mainly protected by fixed-time zero-sequence overcurrent protection, but its ability to detect transition resistance is only about 100 Ω, which is unable to detect single-phase high resistance grounding fault (SPHIF). This paper analyzes [...] Read more.
At present, the small resistance to ground system (SRGS) is mainly protected by fixed-time zero-sequence overcurrent protection, but its ability to detect transition resistance is only about 100 Ω, which is unable to detect single-phase high resistance grounding fault (SPHIF). This paper analyzes the zero-sequence characteristics of SPHIF for SRGS and proposes a SPHIF feeder detection method that uses the current–voltage phase difference. The proposed method is as follows: first, the zero-sequence current phase of each feeder is calculated. Second, the phase voltage root mean square (RMS) value is used to determine the fault phase and obtain its initial phase as the reference value. The introduction of the initial phase of the fault phase voltage can highlight the fault characteristics and improve the sensitivity and reliability of feeder detection, and then CVPD is the difference between each feeder ZSC phase and the reference value. Finally, the magnitude of CVPD is judged. If the CVPD of a particular feeder meets the condition, the feeder is detected as the faulted feeder. Combining the theoretical and practical constraints, the specific adjustment principle and feeder detection logic are given. A large number of simulations show that the proposed method can be successfully detected under the conditions of 5000 Ω transition resistance, –1 dB noise interference, and 40% data missing. Compared with existing methods, the proposed method uses phase voltages that are easy to measure to construct SPHIF feeder detection criteria, without adding additional measurement and communication devices, and can quickly achieve local isolation of SPHIF with better sensitivity, reliability, and immunity to interference. Full article
(This article belongs to the Section Electronic Sensors)
►▼ Show Figures

Figure 1

Figure 1
<p>Diagram of feeder touch tree fault occurred in 10 kV SRGS.</p> Full article ">Figure 2
<p>Zero-sequence equivalent network of SPHIF.</p> Full article ">Figure 3
<p>Simplified zero-sequence networks.</p> Full article ">Figure 4
<p>ZSC and bus ZSV phase volume.</p> Full article ">Figure 5
<p>Fault phase voltage and feeder ZSC phase volume.</p> Full article ">Figure 6
<p>Faulty feeder detection method based on CVPD.</p> Full article ">Figure 7
<p>Extended faulty feeder detection method.</p> Full article ">Figure 8
<p>Voltage vector for a SPHIF in a SRGS.</p> Full article ">Figure 9
<p>SPHIF detection.</p> Full article ">Figure 10
<p>The 10kV small resistance grounding system.</p> Full article ">Figure 11
<p>The CVPD waveform under different transition resistance.</p> Full article ">Figure 12
<p>The CVPD waveform under different fault locations.</p> Full article ">Figure 13
<p>The CVPD waveforms under different initial phase angles of the fault.</p> Full article ">Figure 14
<p>The CVPD waveforms under different feeder lengths.</p> Full article ">Figure 15
<p>Each feeder ZSC waveforms under −1dB noise interference.</p> Full article ">Figure 16
<p>The CVPD waveforms under different noise levels.</p> Full article ">Figure 17
<p>The CVPD waveforms under different data missing ratios.</p> Full article ">
18 pages, 1916 KiB  
Article
Secure IIoT Information Reinforcement Model Based on IIoT Information Platform Using Blockchain
by Yoon-Su Jeong
Sensors 2022, 22(12), 4645; https://doi.org/10.3390/s22124645 - 20 Jun 2022
Cited by 10 | Viewed by 2412
Abstract
Data created at industrial sites through industrial internet of things devices are now being processed automatically or in real-time in the industrial structure, due to the application of artificial intelligence technology to industrial sites. However, the expenses of autonomous or real-time data processing [...] Read more.
Data created at industrial sites through industrial internet of things devices are now being processed automatically or in real-time in the industrial structure, due to the application of artificial intelligence technology to industrial sites. However, the expenses of autonomous or real-time data processing and steady data processing (analysis, prediction, prescription, and implementation) necessitate a new processing method. We propose a blockchain-based industrial internet of things information reinforcement model in this work that may reliably ensure the integrity of industrial internet of things data produced at industrial locations. The proposed model processes industrial internet of things data that may occur at endpoints at industrial sites into the blockchain by processing data generated by the same industrial internet of things device independently. As a result, the IIoT data sent to the industrial internet of things server can be evaluated more readily, and production accuracy may be enhanced. The proposed model optimizes industrial internet of things information linkage by stochastically reflecting the information based on attribute value frequency. By dynamically aggregating the related data of industrial internet of things information acquired as a seed through hierarchical subnets, the proposed model increases stability and accuracy. Furthermore, the proposed model may be used to enhance an organizations’ operational efficiency (consulting and training, for example) and strategic decision-making by utilizing fundamental knowledge about items produced at industrial locations. Furthermore, the proposed model allows for information sharing and system connectivity between industrial locations, allowing for close collaboration between industrial internet of things features. As a result of the performance evaluation, the proposed model included an industrial internet of things sensor to the blockchain, eliminating the need for an extra function in the manufacturing process and reducing the time required to validate the integrity of industrial internet of things data. In addition, as a result of analyzing industrial internet of things data by an algorithm according to the number of simulated clouds, the accuracy of industrial internet of things information was improved by 2.5% to 3%, on average. Full article
(This article belongs to the Special Issue Artificial Intelligence for Security and Privacy in Ad Hoc and Sensor Networks (AI-SPASN))
►▼ Show Figures

Figure 1

Figure 1
<p>Interaction between blockchain and IoT for production sites.</p> Full article ">Figure 2
<p>IIoT information gathering process of proposed model.</p> Full article ">Figure 3
<p>Gathering and sharing of information for platform capabilities.</p> Full article ">Figure 4
<p>Processing of the proposed model.</p> Full article ">Figure 5
<p>Processing of the proposed model.</p> Full article ">Figure 6
<p>Detection of missing values of IIoT information.</p> Full article ">
17 pages, 1214 KiB  
Article
Assessment of Soil Fertility Using Induced Fluorescence and Machine Learning
by Louis Longchamps, Dipankar Mandal and Raj Khosla
Sensors 2022, 22(12), 4644; https://doi.org/10.3390/s22124644 - 20 Jun 2022
Cited by 5 | Viewed by 3168
Abstract
Techniques such as proximal soil sampling are investigated to increase the sampling density and hence the resolution at which nutrient prescription maps are developed. With the advent of a commercial mobile fluorescence sensor, this study assessed the potential of fluorescence to estimate soil [...] Read more.
Techniques such as proximal soil sampling are investigated to increase the sampling density and hence the resolution at which nutrient prescription maps are developed. With the advent of a commercial mobile fluorescence sensor, this study assessed the potential of fluorescence to estimate soil chemical properties and fertilizer recommendations. This experiment was conducted over two years at nine sites on 168 soil samples and used random forest regression to estimate soil properties, fertility classes, and recommended N rates for maize production based on induced fluorescence of air-dried soil samples. Results showed that important soil properties such as soil organic matter, pH, and CEC can be estimated with a correlation of 0.74, 0.75, and 0.75, respectively. When attempting to predict fertility classes, this approach yielded an overall accuracy of 0.54, 0.78, and 0.69 for NO3-N, SOM, and Zn, respectively. The N rate recommendation for maize can be directly estimated by fluorescence readings of the soil with an overall accuracy of 0.78. These results suggest that induced fluorescence is a viable approach for assessing soil fertility. More research is required to transpose these laboratory-acquired soil analysis results to in situ readings successfully. Full article
(This article belongs to the Section Environmental Sensing)
►▼ Show Figures

Figure 1

Figure 1
<p>Multiplex MX3 sensor (<b>a</b>), soil disposed in a plate (container lid) ready for sensing (<b>b</b>), and fluorescence acquisition of the soil sample with the Multiplex MX3 (<b>c</b>).</p> Full article ">Figure 2
<p>(<b>a</b>) Scatter plot of the estimated to observed values of each soil property as per random forest regression analysis. The Pearson’s <span class="html-italic">r</span> coefficient of correlations and two error estimates, i.e., Root Mean Square Error (RMSE) and Mean Absolute Error (MAE), are indicated for both training and test dataset for each soil property. (<b>b</b>) Scatter plot of the estimated to observed values of each soil property as per random forest regression analysis. The Pearson’s <span class="html-italic">r</span> coefficient of correlations and two error estimates, i.e., Root Mean Square Error (RMSE) and Mean Absolute Error (MAE), are indicated for both training and test dataset for each soil property.</p> Full article ">
13 pages, 3321 KiB  
Article
Chlorine Gas Sensor with Surface Temperature Control
by Andrzej Krajewski, Shadi Houshyar, Lijing Wang and Rajiv Padhye
Sensors 2022, 22(12), 4643; https://doi.org/10.3390/s22124643 - 20 Jun 2022
Cited by 1 | Viewed by 2371
Abstract
The work describes the design, manufacturing, and user interface of a thin-film gas transducer platform that is able to provide real-time detection of toxic vapor. This proof-of-concept system has applications in the field of real-time detection of hazardous gaseous agents that are harmful [...] Read more.
The work describes the design, manufacturing, and user interface of a thin-film gas transducer platform that is able to provide real-time detection of toxic vapor. This proof-of-concept system has applications in the field of real-time detection of hazardous gaseous agents that are harmful to the person exposed to the environment. The small-size gas sensor allows for integration with an unmanned aerial vehicle, thus combining high-level mobility with the ability for the real-time detection of hazardous/toxic chemicals or use as a standalone system in industries that deal with harmful gaseous substances. The sensor was designed based on the ability of thin-film metal oxide sensors to detect chlorine gas in real time. Specifically, a concentration of 10 ppm of Cl2 was tested. Full article
(This article belongs to the Special Issue Advances in Nanosensors and Nanogenerators)
►▼ Show Figures

Figure 1

Figure 1
<p>Fabrication process from LTCC ceramic transducer.</p> Full article ">Figure 2
<p>Sensor design.</p> Full article ">Figure 3
<p>LTCC firing process used in the formation of Cl<sub>2</sub> gas sensor transducer.</p> Full article ">Figure 4
<p>Sensor prototypes after firing.</p> Full article ">Figure 5
<p>Heater, sensing area, and PTC resistor connections within the ceramic.</p> Full article ">Figure 6
<p>Feedback-controlled heater solution.</p> Full article ">Figure 7
<p>(<b>a</b>) Temperature vs. power. (<b>b</b>) Feedback resistor’s value variation with power supplied to the heater.</p> Full article ">Figure 8
<p>IR image (<b>top</b>) and the temperature profile (<b>bottom</b>) of the sensor surface with the heater feedback control.</p> Full article ">Figure 9
<p>Test setup for the evaluation of Cl<sub>2</sub> gas–sensitive materials.</p> Full article ">Figure 10
<p>Cl<sub>2</sub> response for In-Sn HN material at 260 °C.</p> Full article ">Figure 11
<p>Cl<sub>2</sub> response for In<sub>2</sub>O<sub>3</sub> NS material at 260 °C.</p> Full article ">Figure 12
<p>Cl<sub>2</sub> response for NiO NP the material at 260 °C.</p> Full article ">
15 pages, 3920 KiB  
Article
Pilot Feasibility Study of a Multi-View Vision Based Scoring Method for Cervical Dystonia
by Chen Ye, Yuhao Xiao, Ruoyu Li, Hongkai Gu, Xinyu Wang, Tianyang Lu and Lingjing Jin
Sensors 2022, 22(12), 4642; https://doi.org/10.3390/s22124642 - 20 Jun 2022
Cited by 2 | Viewed by 3189
Abstract
Abnormal movement of the head and neck is a typical symptom of Cervical Dystonia (CD). Accurate scoring on the severity scale is of great significance for treatment planning. The traditional scoring method is to use a protractor or contact sensors to calculate the [...] Read more.
Abnormal movement of the head and neck is a typical symptom of Cervical Dystonia (CD). Accurate scoring on the severity scale is of great significance for treatment planning. The traditional scoring method is to use a protractor or contact sensors to calculate the angle of the movement, but this method is time-consuming, and it will interfere with the movement of the patient. In the recent outbreak of the coronavirus disease, the need for remote diagnosis and treatment of CD has become extremely urgent for clinical practice. To solve these problems, we propose a multi-view vision based CD severity scale scoring method, which detects the keypoint positions of the patient from the frontal and lateral images, and finally scores the severity scale by calculating head and neck motion angles. We compared the Toronto Western Spasmodic Torticollis Rating Scale (TWSTRS) subscale scores calculated by our vision based method with the scores calculated by a neurologist trained in dyskinesia. An analysis of the correlation coefficient was then conducted. Intra-class correlation (ICC)(3,1) was used to measure absolute accuracy. Our multi-view vision based CD severity scale scoring method demonstrated sufficient validity and reliability. This low-cost and contactless method provides a new potential tool for remote diagnosis and treatment of CD. Full article
(This article belongs to the Topic Human Movement Analysis)
►▼ Show Figures

Figure 1

Figure 1
<p>The diagram of abnormal movement patterns of CD.</p> Full article ">Figure 2
<p>Head positioning cap.</p> Full article ">Figure 3
<p>Devices Diagram for the multi-view vision based method.</p> Full article ">Figure 4
<p>The multi-view vision based method.</p> Full article ">Figure 5
<p>The human keypoints of the subject.</p> Full article ">Figure 6
<p>The scheme of angle calculation.</p> Full article ">Figure 7
<p>The scheme of IMU based method.</p> Full article ">
21 pages, 13115 KiB  
Article
Simulation and Analysis of an FMCW Radar against the UWB EMP Coupling Responses on the Wires
by Kaibai Chen, Shaohua Liu, Min Gao and Xiaodong Zhou
Sensors 2022, 22(12), 4641; https://doi.org/10.3390/s22124641 - 20 Jun 2022
Cited by 4 | Viewed by 3006
Abstract
An ultra-wideband electromagnetic pulse (UWB EMP) can be coupled to an FMCW system through metal wires, causing electronic equipment disturbance or damage. In this paper, a hybrid model is proposed to carry out the interference analysis of UWB EMP coupling responses on the [...] Read more.
An ultra-wideband electromagnetic pulse (UWB EMP) can be coupled to an FMCW system through metal wires, causing electronic equipment disturbance or damage. In this paper, a hybrid model is proposed to carry out the interference analysis of UWB EMP coupling responses on the wires to the FMCW radar. First, a field simulation model of the radar is constructed and the wire coupling responses are calculated. Then, the responses are injected into an FMCW circuit model via data format modification. Finally, we use the FFT transform to identify the spectral peak of the intermediate frequency (IF) output signal, which corresponds to the radar’s detection range. The simulation results show that the type of metal wire has the greatest influence on the amplitude of coupling responses. The spectral peak of the IF output changes to the wrong frequency with the increase of injection power. Applying interference at the end of the circuit can more effectively interfere with the detection of the radar. The investigation provides a theoretical basis for the electromagnetic protection design of the radar. Full article
(This article belongs to the Special Issue Radar Sensors for Target Tracking and Localization)
►▼ Show Figures

Figure 1

Figure 1
<p>The brief ranging principle of the FMCW radar.</p> Full article ">Figure 2
<p>The flowchart of the proposed hybrid model.</p> Full article ">Figure 3
<p>The external structure of the radar.</p> Full article ">Figure 4
<p>The internal structure of the radar.</p> Full article ">Figure 5
<p>The external structure of the field model.</p> Full article ">Figure 6
<p>The internal structure of the field model.</p> Full article ">Figure 7
<p>The incident direction of UWB EMP.</p> Full article ">Figure 8
<p>The composition of the circuit model of the FMCW radar.</p> Full article ">Figure 9
<p>IF output spectrum.</p> Full article ">Figure 10
<p>Time-domain waveform and the frequency-domain waveform of UWB EMP. (<b>a</b>) Time-domain waveform of UWB EMP; (<b>b</b>) frequency-domain waveform of UWB EMP.</p> Full article ">Figure 11
<p>The actual geometry of the wires. (<b>a</b>) A single wire; (<b>b</b>) A twisted wire; (<b>c</b>) A coaxial wire.</p> Full article ">Figure 12
<p>The structure of electromagnetic field.</p> Full article ">Figure 13
<p>The responses of different categories of metal wire.</p> Full article ">Figure 14
<p>The responses of different lengths of single wire.</p> Full article ">Figure 15
<p>Relationship between peak time and peak voltage with the wire lengths.</p> Full article ">Figure 16
<p>The responses of the different radii of a single wire.</p> Full article ">Figure 17
<p>The geometry of the wires.</p> Full article ">Figure 18
<p>The responses of the different curvatures of a single wire.</p> Full article ">Figure 19
<p>The geometry of the wires.</p> Full article ">Figure 20
<p>The responses of the different number of wires.</p> Full article ">Figure 21
<p>The geometry of the wires.</p> Full article ">Figure 22
<p>The responses of different distances among the wires.</p> Full article ">Figure 23
<p>The schematic of the circuit analysis.</p> Full article ">Figure 24
<p>The injection interference law in node 5.</p> Full article ">Figure 25
<p>The injection interference law in node 6.</p> Full article ">Figure 26
<p>The injection interference law in nodes 5 and 6.</p> Full article ">
14 pages, 1390 KiB  
Article
FDG-PET to T1 Weighted MRI Translation with 3D Elicit Generative Adversarial Network (E-GAN)
by Farideh Bazangani, Frédéric J. P. Richard, Badih Ghattas and Eric Guedj
Sensors 2022, 22(12), 4640; https://doi.org/10.3390/s22124640 - 20 Jun 2022
Cited by 10 | Viewed by 3363
Abstract
Objective: With the strengths of deep learning, computer-aided diagnosis (CAD) is a hot topic for researchers in medical image analysis. One of the main requirements for training a deep learning model is providing enough data for the network. However, in medical images, due [...] Read more.
Objective: With the strengths of deep learning, computer-aided diagnosis (CAD) is a hot topic for researchers in medical image analysis. One of the main requirements for training a deep learning model is providing enough data for the network. However, in medical images, due to the difficulties of data collection and data privacy, finding an appropriate dataset (balanced, enough samples, etc.) is quite a challenge. Although image synthesis could be beneficial to overcome this issue, synthesizing 3D images is a hard task. The main objective of this paper is to generate 3D T1 weighted MRI corresponding to FDG-PET. In this study, we propose a separable convolution-based Elicit generative adversarial network (E-GAN). The proposed architecture can reconstruct 3D T1 weighted MRI from 2D high-level features and geometrical information retrieved from a Sobel filter. Experimental results on the ADNI datasets for healthy subjects show that the proposed model improves the quality of images compared with the state of the art. In addition, the evaluation of E-GAN and the state of art methods gives a better result on the structural information (13.73% improvement for PSNR and 22.95% for SSIM compared to Pix2Pix GAN) and textural information (6.9% improvements for homogeneity error in Haralick features compared to Pix2Pix GAN). Full article
(This article belongs to the Special Issue Machine Learning Techniques for Medical Imaging, Sensing, and Analysis)
►▼ Show Figures

Figure 1

Figure 1
<p>Network architecture of E-GAN for image-to-image translation.</p> Full article ">Figure 2
<p>Configuration of the Elicit Network for projection the features in 2D.</p> Full article ">Figure 3
<p>3D Sobel operation in x, y and z.</p> Full article ">Figure 4
<p>A pair of FDG-PET (left image), the ground truth T1 weighted MRI (middle image), and the generated image (right image) with E-GAN.</p> Full article ">Figure 5
<p>Experimental results for translating PET images to corresponding MRI with 3D DCGAN [<a href="#B35-sensors-22-04640" class="html-bibr">35</a>], 3D WGAN [<a href="#B25-sensors-22-04640" class="html-bibr">25</a>], 3D Pix2Pix GAN [<a href="#B17-sensors-22-04640" class="html-bibr">17</a>], and the proposed method.</p> Full article ">Figure 6
<p>Synthetic MRI with the Sobel filter (<b>a</b>) and without the Sobel filter (<b>b</b>).</p> Full article ">Figure 7
<p>The ground truth (<b>a</b>) and the synthetic T1 weighted MRI (<b>b</b>).</p> Full article ">Figure 8
<p>The Average voxel values in the translated image and the ground truth for 10 subjects.</p> Full article ">Figure 9
<p>The Co-occurrence matrix DCGAN [<a href="#B35-sensors-22-04640" class="html-bibr">35</a>] (<b>A</b>), WGAN [<a href="#B25-sensors-22-04640" class="html-bibr">25</a>] (<b>B</b>), Pix2Pix GAN [<a href="#B17-sensors-22-04640" class="html-bibr">17</a>] (<b>C</b>), Proposed model (<b>D</b>) and the ground truth (<b>E</b>) in 3D.</p> Full article ">
17 pages, 3199 KiB  
Article
Severity Estimation for Interturn Short-Circuit and Demagnetization Faults through Self-Attention Network
by Hojin Lee, Hyeyun Jeong, Seongyun Kim and Sang Woo Kim
Sensors 2022, 22(12), 4639; https://doi.org/10.3390/s22124639 - 20 Jun 2022
Cited by 4 | Viewed by 2200
Abstract
This study presents a novel interturn short-circuit fault (ISCF) and demagnetization fault (DF) diagnosis strategy based on a self-attention-based severity estimation network (SASEN). We analyze the effects of the ISCF and DF in a permanent-magnet synchronous machine and select appropriate inputs for estimating [...] Read more.
This study presents a novel interturn short-circuit fault (ISCF) and demagnetization fault (DF) diagnosis strategy based on a self-attention-based severity estimation network (SASEN). We analyze the effects of the ISCF and DF in a permanent-magnet synchronous machine and select appropriate inputs for estimating the fault severities, i.e., a positive-sequence voltage and current and negative-sequence voltage and current. The chosen inputs are fed into the SASEN to estimate fault indicators for quantifying the fault severities of the ISCF and DF. The SASEN comprises an encoder and decoder based on a self-attention module. The self-attention mechanism enhances the high-dimensional feature extraction and regression ability of the network by concentrating on specific sequence representations, thereby supporting the estimation of the fault severities. The proposed strategy can diagnose a hybrid fault in which the ISCF and DF occur simultaneously and does not require the exact model and parameters essential for the existing method for estimating the fault severity. The effectiveness and feasibility of the proposed fault diagnosis strategy are demonstrated through experimental results based on various fault cases and load torque conditions. Full article
(This article belongs to the Collection Artificial Intelligence for Data-Driven Fault Detection and Diagnosis)
►▼ Show Figures

Figure 1

Figure 1
<p>Structure of self-attention module.</p> Full article ">Figure 2
<p>Overall architecture of the self-attention-based severity estimation network (SASEN).</p> Full article ">Figure 3
<p>Overall structure of fault diagnosis system.</p> Full article ">Figure 4
<p>Experimental setup of the interior permanent-magnet synchronous machine (IPMSM).</p> Full article ">Figure 5
<p>Demagnetized IPMSM. Permanent magnets are replaced with dummies.</p> Full article ">Figure 6
<p>Change in the load torque for the test 1.</p> Full article ">Figure 7
<p>Test results for the interturn short-circuit fault (ISCF) at transient load torque. (<b>a</b>) Fault indicator for the ISCF. (<b>b</b>) Fault indicator for the demagnetization fault (DF). The solid line and dashed line represent the estimated and real fault indicators, respectively.</p> Full article ">Figure 8
<p>Test results for the DF at transient load torque. (<b>a</b>) Fault indicator for the ISCF. (<b>b</b>) Fault indicator for the DF. The solid line and dashed line represent the estimated and real fault indicators, respectively.</p> Full article ">Figure 9
<p>Test results for the hybrid fault (HF) at transient load torque. (<b>a</b>) Fault indicator for the ISCF. (<b>b</b>) Fault indicator for the DF. The solid line and dashed line represent the estimated and real fault indicators, respectively.</p> Full article ">Figure 10
<p>Test results for the untrained load torque. Test case is HF5 at untrained load torques of 1.2 to 3.8 Nm. (<b>a</b>) Fault indicator for the ISCF. (<b>b</b>) Fault indicator for the DF.</p> Full article ">Figure 10 Cont.
<p>Test results for the untrained load torque. Test case is HF5 at untrained load torques of 1.2 to 3.8 Nm. (<b>a</b>) Fault indicator for the ISCF. (<b>b</b>) Fault indicator for the DF.</p> Full article ">
14 pages, 2511 KiB  
Article
Biosensors for Klebsiella pneumoniae with Molecularly Imprinted Polymer (MIP) Technique
by Chuchart Pintavirooj, Naphatsawan Vongmanee, Wannisa Sukjee, Chak Sangma and Sarinporn Visitsattapongse
Sensors 2022, 22(12), 4638; https://doi.org/10.3390/s22124638 - 20 Jun 2022
Cited by 14 | Viewed by 3383
Abstract
Nosocomial infection is one of the most important problems that occurs in hospitals, as it directly affects susceptible patients or patients with immune deficiency. Klebsiella pneumoniae (K. pneumoniae) is the most common cause of nosocomial infections in hospitals. K. pneumoniae can cause [...] Read more.
Nosocomial infection is one of the most important problems that occurs in hospitals, as it directly affects susceptible patients or patients with immune deficiency. Klebsiella pneumoniae (K. pneumoniae) is the most common cause of nosocomial infections in hospitals. K. pneumoniae can cause various diseases such as pneumonia, urinary tract infections, septicemias, and soft tissue infections, and it has also become highly resistant to antibiotics. The principal routes for the transmission of K. pneumoniae are via the gastrointestinal tract and the hands of hospital personnel via healthcare workers, patients, hospital equipment, and interventional procedures. These bacteria can spread rapidly in the hospital environment and tend to cause nosocomial outbreaks. In this research, we developed a MIP-based electrochemical biosensor to detect K. pneumoniae. Quantitative detection was performed using an electrochemical technique to measure the changes in electrical signals in different concentrations of K. pneumoniae ranging from 10 to 105 CFU/mL. Our MIP-based K. pneumoniae sensor was found to achieve a high linear response, with an R2 value of 0.9919. A sensitivity test was also performed on bacteria with a similar structure to that of K. pneumoniae. The sensitivity results show that the MIP-based K. pneumoniae biosensor with a gold electrode was the most sensitive, with a 7.51 (% relative current/log concentration) when compared with the MIP sensor applied with Pseudomonas aeruginosa and Enterococcus faecalis, where the sensitivity was 2.634 and 2.226, respectively. Our sensor was also able to achieve a limit of detection (LOD) of 0.012 CFU/mL and limit of quantitation (LOQ) of 1.61 CFU/mL. Full article
(This article belongs to the Section Biosensors)
►▼ Show Figures

Figure 1

Figure 1
<p>Schematic representation of <span class="html-italic">K. pneumoniae</span> on LB agar and the serial dilution method.</p> Full article ">Figure 2
<p>Schematic representation of the preparation of polymer-GO on gold electrode for <span class="html-italic">K. pneumoniae</span> detection.</p> Full article ">Figure 3
<p>(<b>a</b>) Magnified SEM image at 10,000×; (<b>b</b>) the surface of whole <span class="html-italic">K. pneumoniae</span> on the SPE, with a size of approximately 0.5 × 1.7 µm (SEM image at 50,000×).</p> Full article ">Figure 4
<p>Comparison of linearity range for conditions 1 to 10 on the carbon electrode.</p> Full article ">Figure 5
<p>(<b>a</b>) Cyclic voltammogram of <span class="html-italic">K. pneumoniae</span> at concentration levels in condition 5 on gold electrode; (<b>b</b>) cyclic voltammogram of <span class="html-italic">K. pneumoniae</span> at concentration levels in condition 9 on gold electrode.</p> Full article ">Figure 6
<p>Comparison of linearity range for condition 5 and condition 9 on the gold electrode (* is referred to multiplication).</p> Full article ">Figure 7
<p>The plot of current changes as a linear function of log <span class="html-italic">K. pneumoniae</span> concentration gains from MIP composite on the gold electrode compared with the results from the control experiments using <span class="html-italic">P. aeruginosa</span> and <span class="html-italic">E. faecalis.</span> The number of replicated measurements is 3. (* is referred to multiplication).</p> Full article ">
18 pages, 5805 KiB  
Article
Method for Continuous Integration and Deployment Using a Pipeline Generator for Agile Software Projects
by Ionut-Catalin Donca, Ovidiu Petru Stan, Marius Misaros, Dan Gota and Liviu Miclea
Sensors 2022, 22(12), 4637; https://doi.org/10.3390/s22124637 - 20 Jun 2022
Cited by 20 | Viewed by 6767
Abstract
Lately, the software development industry is going through a slow but real transformation. Software is increasingly a part of everything, and, software developers, are trying to cope with this exploding demand through more automation. The pipelining technique of continuous integration (CI) and continuous [...] Read more.
Lately, the software development industry is going through a slow but real transformation. Software is increasingly a part of everything, and, software developers, are trying to cope with this exploding demand through more automation. The pipelining technique of continuous integration (CI) and continuous delivery (CD) has developed considerably due to the overwhelming demand for the deployment and deliverability of new features and applications. As a result, DevOps approaches and Agile principles have been developed, in which developers collaborate closely with infrastructure engineers to guarantee that their applications are deployed quickly and reliably. Thanks to pipeline approach thinking, the efficiency of projects has greatly improved. Agile practices represent the introduction to the system of new features in each sprint delivery. Those practices may contain well-developed features or can contain bugs or failures which impact the delivery. The pipeline approach, depicted in this paper, overcomes the problems of delivery, improving the delivery timeline, the test load steps, and the benchmarking tasks. It decreases system interruption by integrating multiple test steps and adds stability and deliverability to the entire process. It provides standardization which means having an established, time-tested process to use, and can also decrease ambiguity and guesswork, guarantee quality and boost productivity. This tool is developed with an interpreted language, namely Bash, which offers an easier method to integrate it into any platform. Based on the experimental results, we demonstrate the value that this solution currently creates. This solution provides an effective and efficient way to generate, manage, customize, and automate Agile-based CI and CD projects through automated pipelines. The suggested system acts as a starting point for standard CI/CD processes, caches Docker layers for subsequent usage, and implements highly available deliverables in a Kubernetes cluster using Helm. Changing the principles of this solution and expanding it into multiple platforms (windows) will be addressed in a future discussion. Full article
(This article belongs to the Special Issue Intelligent Control and Testing Systems and Applications)
►▼ Show Figures

Figure 1

Figure 1
<p>UML Diagram of the proposed solution.</p> Full article ">Figure 2
<p>Extended UML Diagram of the proposed solution.</p> Full article ">Figure 3
<p>Pipeline generator code snippet.</p> Full article ">Figure 4
<p>Versioning flow described.</p> Full article ">Figure 5
<p>Versioning code block.</p> Full article ">Figure 6
<p>Pipeline diagram flow.</p> Full article ">Figure 7
<p>The pipeline diagram flow dependencies.</p> Full article ">Figure 8
<p>Build image function.</p> Full article ">Figure 9
<p>Docker-cached layers dataset.</p> Full article ">Figure 10
<p>An overall proposed solution flowchart diagram.</p> Full article ">Figure 11
<p>Helm installation/upgrade process.</p> Full article ">Figure 12
<p>Pipeline steps and duration.</p> Full article ">Figure 13
<p>Pipeline type time difference in seconds.</p> Full article ">Figure 14
<p>Pipeline type infrastructure costs.</p> Full article ">Figure 15
<p>RabbitMQ deployed with the proposed solution.</p> Full article ">
13 pages, 3409 KiB  
Article
Thermo-Optical Sensitivity of Whispering Gallery Modes in As2S3 Chalcogenide Glass Microresonators
by Alexey V. Andrianov, Maria P. Marisova and Elena A. Anashkina
Sensors 2022, 22(12), 4636; https://doi.org/10.3390/s22124636 - 20 Jun 2022
Cited by 10 | Viewed by 2559
Abstract
Glass microresonators with whispering gallery modes (WGMs) have a lot of diversified applications, including applications for sensing based on thermo-optical effects. Chalcogenide glass microresonators have a noticeably higher temperature sensitivity compared to silica ones, but only a few works have been devoted to [...] Read more.
Glass microresonators with whispering gallery modes (WGMs) have a lot of diversified applications, including applications for sensing based on thermo-optical effects. Chalcogenide glass microresonators have a noticeably higher temperature sensitivity compared to silica ones, but only a few works have been devoted to the study of their thermo-optical properties. We present experimental and theoretical studies of thermo-optical effects in microspheres made of an As2S3 chalcogenide glass fiber. We investigated the steady-state and transient temperature distributions caused by heating due to the partial thermalization of the pump power and found the corresponding wavelength shifts of the WGMs. The experimental measurements of the thermal response time, thermo-optical shifts of the WGMs, and heat power sensitivity in microspheres with diameters of 80–380 µm are in a good agreement with the theoretically predicted dependences. The calculated temperature sensitivity of 42 pm/K does not depend on diameter for microspheres made of commercially available chalcogenide fiber, which may play an important role in the development of temperature sensors. Full article
(This article belongs to the Special Issue State-of-the-Art Optical Sensors Technology in Russia 2021-2022)
►▼ Show Figures

Figure 1

Figure 1
<p>(<b>a</b>) Simplified scheme of the experimental setup; CW–continuous wave, ASE–amplified spontaneous emission, OSA–optical spectrum analyzer, PD–photodetector. (<b>b</b>) Images of experimental As<sub>2</sub>S<sub>3</sub> chalcogenide glass microspheres obtained with optical microscope (with different magnification). (<b>c</b>) Measured resonance dip and its Lorentz fit demonstrating Q-factor.</p> Full article ">Figure 2
<p>(<b>a</b>) Scheme of microresonator geometry used in simulations. (<b>b</b>) Effective mode area (red curve, left axis) and effective mode volume (brown curve, right axis) calculated at a wavelength of ~1.55 μm.</p> Full article ">Figure 3
<p>Numerical results. (<b>a</b>) Distribution of steady-state temperature increase in microresonator with diameter <span class="html-italic">d</span> = 140 µm for thermalized power <span class="html-italic">P</span> = 1 mW. (<b>b</b>) Temporal dynamics of temperature increases averaged over fundamental mode (Δ<span class="html-italic">T</span><sub>mode</sub>) and over microsphere (Δ<span class="html-italic">T</span><sub>av</sub>) when pump is switched on with a thermalized power of 1 mW and switched off at the moment <span class="html-italic">t</span><sub>off</sub>, marked by the vertical gray dashed line. ‘Fit’ is approximation of Δ<span class="html-italic">T</span><sub>av</sub> by exponential decay with characteristic time <span class="html-italic">t</span><sub>0</sub> when pump is switched off. (<b>c</b>) Dependence of <span class="html-italic">t</span><sub>0</sub> on microsphere diameter obtained by direct simulations and fitted by <span class="html-italic">t</span><sub>0</sub> = <span class="html-italic">C</span><sub>1</sub>⸱<span class="html-italic">d</span><sup>2</sup>, where <span class="html-italic">C</span><sub>1</sub> = 5.2 × 10<sup>−6</sup> s⸱µm<sup>−2</sup>. Steady-state temperature increase averaged over the microsphere (<b>d</b>) and over the fundamental mode (<b>e</b>) as functions of thermalized power and microsphere diameter. (<b>f</b>) Average temperature increases for <span class="html-italic">d</span> = 140 µm and their linear fits (Δ<span class="html-italic">T</span>~<span class="html-italic">P</span>). (<b>g</b>) Average temperature increases for <span class="html-italic">P</span> = 1 mW and their fits (Δ<span class="html-italic">T</span>~1/<span class="html-italic">d</span>).</p> Full article ">Figure 4
<p>Numerically simulated steady-state sensitivity Δ<span class="html-italic">λ/</span>Δ<span class="html-italic">T</span> as a function of <span class="html-italic">d</span>.</p> Full article ">Figure 5
<p>Numerical results. (<b>a</b>) Steady-state thermo-optical shift of WGMs, Δλ, as a function of thermalized power and microsphere diameter. (<b>b</b>) Δ<span class="html-italic">λ</span> as a function of <span class="html-italic">P</span> for <span class="html-italic">d</span> = 140 µm and its linear fit. (<b>c</b>) Δ<span class="html-italic">λ</span> as a function of <span class="html-italic">d</span> for <span class="html-italic">P</span> = 1 mW and its fit ~1/<span class="html-italic">d</span>.</p> Full article ">Figure 6
<p>Experimental (<b>a</b>) and numerically simulated (<b>b</b>) dependences of thermo-optical shift in WGMs Δλ at time <span class="html-italic">t</span>. Thermalized power is 0.9 mW for <span class="html-italic">t</span> &lt; 0 and the pump is switched off at <span class="html-italic">t</span> = 0. ‘Fit’ is approximation of Δλ by exponential decay with characteristic time <span class="html-italic">t</span><sub>Δλ</sub> for <span class="html-italic">t</span> &gt; 0; (<b>c</b>) <span class="html-italic">t</span><sub>Δλ</sub> dependence on microsphere diameter obtained by experimental measurements, by direct simulations and fitted by <span class="html-italic">t</span><sub>Δλ</sub> = <span class="html-italic">C</span><sub>2</sub>⸱<span class="html-italic">d</span><sup>2</sup>, where <span class="html-italic">C</span><sub>2</sub> = 4.6 × 10<sup>−6</sup> s⸱µm<sup>−2</sup>.</p> Full article ">Figure 7
<p>Heat power sensitivity Δ<span class="html-italic">λ</span>/<span class="html-italic">P</span> versus microsphere diameter obtained by experimental measurements, by direct simulations, and fitted by Δ<span class="html-italic">λ</span>/<span class="html-italic">P</span> = <span class="html-italic">C</span><sub>3</sub>/<span class="html-italic">d</span>, where <span class="html-italic">C</span><sub>3</sub> = 230 nm⸱µm/mW.</p> Full article ">
13 pages, 3282 KiB  
Article
Three-Dimensional (3D) Imaging Technology to Monitor Growth and Development of Holstein Heifers and Estimate Body Weight, a Preliminary Study
by Yannick Le Cozler, Elodie Brachet, Laurianne Bourguignon, Laurent Delattre, Thibaut Luginbuhl and Philippe Faverdin
Sensors 2022, 22(12), 4635; https://doi.org/10.3390/s22124635 - 19 Jun 2022
Cited by 7 | Viewed by 2421
Abstract
The choice of rearing strategy for dairy cows can have an effect on production yield, at least during the first lactation. For this reason, it is important to closely monitor the growth and development of young heifers. Unfortunately, current methods for evaluation can [...] Read more.
The choice of rearing strategy for dairy cows can have an effect on production yield, at least during the first lactation. For this reason, it is important to closely monitor the growth and development of young heifers. Unfortunately, current methods for evaluation can be costly, time-consuming, and dangerous because of the need to physically manipulate animals, and as a result, this type of monitoring is seldom performed on farms. One potential solution may be the use of tools based on three-dimensional (3D) imaging, which has been studied in adult cows but not yet in growing individuals. In this study, an imaging approach that was previously validated for adult cows was tested on a pilot population of five randomly selected growing Holstein heifers, from 5 weeks of age to the end of the first gestation. Once a month, all heifers were weighed and an individual 3D image was recorded. From these images, we estimated growth trends in morphological traits such as heart girth or withers height (188.1 ± 3.7 cm and 133.5 ± 6.0 cm on average at one year of age, respectively). From other traits, such as body surface area and volume (5.21 ± 0.32 m2 and 0.43 ± 0.05 m3 on average at one year of age, respectively), we estimated body weight based on volume (402.4 ± 37.5 kg at one year of age). Body weight estimates from images were on average 9.7% higher than values recorded by the weighing scale (366.8 ± 47.2 kg), but this difference varied with age (19.1% and 1.8% at 6 and 20 months of age, respectively). To increase accuracy, the predictive model developed for adult cows was adapted and completed with complementary data on young heifers. Using imaging data, it was also possible to analyze changes in the surface-to-volume ratio that occurred as body weight and age increased. In sum, 3D imaging technology is an easy-to-use tool for following the growth and management of heifers and should become increasingly accurate as more data are collected on this population. Full article
(This article belongs to the Section Smart Agriculture)
►▼ Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Abnormal image representing the presence of a “skirt” on the legs. When the number of points is insufficient, the software fills in the volume between the points located on different legs (which is not the case when there are enough points).</p> Full article ">Figure 2
<p>Dynamics of body weight (<b>a</b>), heart girth (<b>b</b>), wither height (<b>c</b>), surface area (<b>d</b>), and volume (<b>e</b>) as a function of age, and (<b>f</b>) surface-to-volume ratio as a function of body weight for five Holstein heifers, from 5 weeks of age until the end of gestation.</p> Full article ">Figure 3
<p>Measured values of body weight (BW) compared to values estimated (BWe) as a function of body surface area (BSA), volume, withers height (WH), and/or buttock width (WH), using equations from Elting [<a href="#B10-sensors-22-04635" class="html-bibr">10</a>] (<b>a</b>,<b>b</b>); or Le Cozler et al. [<a href="#B13-sensors-22-04635" class="html-bibr">13</a>], (<b>c</b>–<b>e</b>).</p> Full article ">Figure 3 Cont.
<p>Measured values of body weight (BW) compared to values estimated (BWe) as a function of body surface area (BSA), volume, withers height (WH), and/or buttock width (WH), using equations from Elting [<a href="#B10-sensors-22-04635" class="html-bibr">10</a>] (<b>a</b>,<b>b</b>); or Le Cozler et al. [<a href="#B13-sensors-22-04635" class="html-bibr">13</a>], (<b>c</b>–<b>e</b>).</p> Full article ">Figure 4
<p>Changes in surface-to-volume ratio as a function of body weight for five Holstein heifers, from 5 weeks of age until the end of gestation (empty circles; <span class="html-italic">n</span> = 68 observations), supplemented with data from adult cows (full circles; <span class="html-italic">n</span> = 177 observations).</p> Full article ">
17 pages, 3680 KiB  
Article
Micro-Expression-Based Emotion Recognition Using Waterfall Atrous Spatial Pyramid Pooling Networks
by Marzuraikah Mohd Stofa, Mohd Asyraf Zulkifley and Muhammad Ammirrul Atiqi Mohd Zainuri
Sensors 2022, 22(12), 4634; https://doi.org/10.3390/s22124634 - 19 Jun 2022
Cited by 10 | Viewed by 2753
Abstract
Understanding a person’s attitude or sentiment from their facial expressions has long been a straightforward task for humans. Numerous methods and techniques have been used to classify and interpret human emotions that are commonly communicated through facial expressions, with either macro- or micro-expressions. [...] Read more.
Understanding a person’s attitude or sentiment from their facial expressions has long been a straightforward task for humans. Numerous methods and techniques have been used to classify and interpret human emotions that are commonly communicated through facial expressions, with either macro- or micro-expressions. However, performing this task using computer-based techniques or algorithms has been proven to be extremely difficult, whereby it is a time-consuming task to annotate it manually. Compared to macro-expressions, micro-expressions manifest the real emotional cues of a human, which they try to suppress and hide. Different methods and algorithms for recognizing emotions using micro-expressions are examined in this research, and the results are presented in a comparative approach. The proposed technique is based on a multi-scale deep learning approach that aims to extract facial cues of various subjects under various conditions. Then, two popular multi-scale approaches are explored, Spatial Pyramid Pooling (SPP) and Atrous Spatial Pyramid Pooling (ASPP), which are then optimized to suit the purpose of emotion recognition using micro-expression cues. There are four new architectures introduced in this paper based on multi-layer multi-scale convolutional networks using both direct and waterfall network flows. The experimental results show that the ASPP module with waterfall network flow, which we coined as WASPP-Net, outperforms the state-of-the-art benchmark techniques with an accuracy of 80.5%. For future work, a high-resolution approach to multi-scale approaches can be explored to further improve the recognition performance. Full article
(This article belongs to the Topic Applications in Image Analysis and Pattern Recognition)
►▼ Show Figures

Figure 1

Figure 1
<p>Differences in facial muscle movement for happy emotion among the test subjects: (<b>a</b>) subject 1; (<b>b</b>) subject 2; (<b>c</b>) subject 3.</p> Full article ">Figure 2
<p>Basic SPP module architecture.</p> Full article ">Figure 3
<p>Two placement strategies of the SPP module in the base CNN model.</p> Full article ">Figure 4
<p>Basic ASPP module architecture.</p> Full article ">Figure 5
<p>Two placement strategies of the ASPP module in the base CNN model.</p> Full article ">Figure 6
<p>Direct network flow of the SPP and ASPP modules: (<b>a</b>) DSPP-Net architecture; (<b>b</b>) DASPP-Net architecture.</p> Full article ">Figure 7
<p>Waterfall network flow for SPP and ASPP modules: (<b>a</b>) WSPP-Net architecture; (<b>b</b>) WASPP-Net architecture.</p> Full article ">Figure 8
<p>The training graph performance: (<b>a</b>) DSPP-Net architecture; (<b>b</b>) WSPP-Net architecture.</p> Full article ">Figure 9
<p>The training graph performance: (<b>a</b>) DASPP-Net architecture; (<b>b</b>) WASPP-Net architecture.</p> Full article ">
22 pages, 8960 KiB  
Article
Emotion Recognition for Partial Faces Using a Feature Vector Technique
by Ratanak Khoeun, Ponlawat Chophuk and Krisana Chinnasarn
Sensors 2022, 22(12), 4633; https://doi.org/10.3390/s22124633 - 19 Jun 2022
Cited by 9 | Viewed by 3264
Abstract
Wearing a facial mask is indispensable in the COVID-19 pandemic; however, it has tremendous effects on the performance of existing facial emotion recognition approaches. In this paper, we propose a feature vector technique comprising three main steps to recognize emotions from facial mask [...] Read more.
Wearing a facial mask is indispensable in the COVID-19 pandemic; however, it has tremendous effects on the performance of existing facial emotion recognition approaches. In this paper, we propose a feature vector technique comprising three main steps to recognize emotions from facial mask images. First, a synthetic mask is used to cover the facial input image. With only the upper part of the image showing, and including only the eyes, eyebrows, a portion of the bridge of the nose, and the forehead, the boundary and regional representation technique is applied. Second, a feature extraction technique based on our proposed rapid landmark detection method employing the infinity shape is utilized to flexibly extract a set of feature vectors that can effectively indicate the characteristics of the partially occluded masked face. Finally, those features, including the location of the detected landmarks and the Histograms of the Oriented Gradients, are brought into the classification process by adopting CNN and LSTM; the experimental results are then evaluated using images from the CK+ and RAF-DB data sets. As the result, our proposed method outperforms existing cutting-edge approaches and demonstrates better performance, achieving 99.30% and 95.58% accuracy on CK+ and RAF-DB, respectively. Full article
(This article belongs to the Topic Data Analytics and Machine Learning in Artificial Emotional Intelligence)
►▼ Show Figures

Figure 1

Figure 1
<p>Upper portion of the face and AU samples of the baseline emotions; (<b>a</b>) upper face; (<b>b</b>) AUs of each emotion; (<b>c</b>) comparison of the upper AUs of each pair of emotions.</p> Full article ">Figure 2
<p>Traditional landmarks.</p> Full article ">Figure 3
<p>Histograms of the distance between landmarks across all emotions from nine CK+ samples.</p> Full article ">Figure 4
<p>(<b>a</b>) Histograms of the distance between landmarks across all emotions from all of the samples in CK+; (<b>b</b>) graph of the sorted distances of all of the samples in CK+.</p> Full article ">Figure 5
<p>Each traditional upper landmark has a similar location to one another across the different emotions of a person; <math display="inline"><semantics> <mrow> <mi>N</mi> <mi>e</mi> <mi>u</mi> <mi>t</mi> <mi>r</mi> <mi>a</mi> <msub> <mi>l</mi> <mrow> <mn>1</mn> <mo>,</mo> <mn>17</mn> <mo>,</mo> <mn>18</mn> <mo>,</mo> <mo>…</mo> <mo>,</mo> <mn>29</mn> </mrow> </msub> <mo>≅</mo> <mi>F</mi> <mi>e</mi> <mi>a</mi> <msub> <mi>r</mi> <mrow> <mn>1</mn> <mo>,</mo> <mn>17</mn> <mo>,</mo> <mn>18</mn> <mo>,</mo> <mo>…</mo> <mo>,</mo> <mn>29</mn> </mrow> </msub> <mo>≅</mo> <mi>S</mi> <mi>u</mi> <mi>r</mi> <mi>p</mi> <mi>r</mi> <mi>i</mi> <mi>s</mi> <msub> <mi>e</mi> <mrow> <mn>1</mn> <mo>,</mo> <mn>17</mn> <mo>,</mo> <mn>18</mn> <mo>,</mo> <mo>…</mo> <mo>,</mo> <mn>29</mn> </mrow> </msub> <mo>≅</mo> <mi>S</mi> <mi>a</mi> <msub> <mi>d</mi> <mrow> <mn>1</mn> <mo>,</mo> <mn>17</mn> <mo>,</mo> <mn>18</mn> <mo>,</mo> <mo>…</mo> <mo>,</mo> <mn>29</mn> </mrow> </msub> <mo>≅</mo> <mi>A</mi> <mi>n</mi> <mi>g</mi> <mi>e</mi> <msub> <mi>r</mi> <mrow> <mn>1</mn> <mo>,</mo> <mn>17</mn> <mo>,</mo> <mn>18</mn> <mo>,</mo> <mo>…</mo> <mo>,</mo> <mn>29</mn> </mrow> </msub> <mo>≅</mo> <mi>D</mi> <mi>i</mi> <mi>s</mi> <mi>g</mi> <mi>u</mi> <mi>s</mi> <msub> <mi>t</mi> <mrow> <mn>1</mn> <mo>,</mo> <mn>17</mn> <mo>,</mo> <mn>18</mn> <mo>,</mo> <mo>…</mo> <mo>,</mo> <mn>29</mn> </mrow> </msub> <mo>≅</mo> <mi>H</mi> <mi>a</mi> <mi>p</mi> <mi>p</mi> <mi>i</mi> <mi>n</mi> <mi>e</mi> <mi>s</mi> <msub> <mi>s</mi> <mrow> <mn>1</mn> <mo>,</mo> <mn>17</mn> <mo>,</mo> <mn>18</mn> <mo>,</mo> <mo>…</mo> <mo>,</mo> <mn>29</mn> </mrow> </msub> <mo>≅</mo> <mi>C</mi> <mi>o</mi> <mi>n</mi> <mi>t</mi> <mi>e</mi> <mi>m</mi> <mi>p</mi> <msub> <mi>t</mi> <mrow> <mn>1</mn> <mo>,</mo> <mn>17</mn> <mo>,</mo> <mn>18</mn> <mo>,</mo> <mo>…</mo> <mo>,</mo> <mn>29</mn> </mrow> </msub> </mrow> </semantics></math>.</p> Full article ">Figure 6
<p>Landmarks detected using our proposed method.</p> Full article ">Figure 7
<p>Illustration of infinity shapes.</p> Full article ">Figure 8
<p>Illustration of the proposed method and the algorithm.</p> Full article ">Figure 9
<p>Upper face detection; (<b>a</b>) CK+ original image; (<b>b</b>) face detected using Dlib with the overlapping synthetic facial mask; (<b>c</b>) detected upper face.</p> Full article ">Figure 10
<p>Detected and undetected landmarks.</p> Full article ">Figure 11
<p>Example of the HOG features of each detected landmark, the HOG feature vector, and the algorithm.</p> Full article ">Figure 12
<p>(<b>a</b>) LSTM components; (<b>b</b>) machine learning architecture.</p> Full article ">Figure 13
<p>Results of our proposed method, including the confusion matrix as well as the loss and accuracy graphs; (<b>a</b>) results for CK+; (<b>b</b>) results for RAF-DB.</p> Full article ">Figure 14
<p>Comparison of the accuracy metrics of the proposed method used on the CK+ and RAF-DB data sets.</p> Full article ">Figure 15
<p>All of the detected landmarks for eight emotions in the CK+ data set; (<b>a</b>) images of upper faces with the infinity shape and detected landmarks; (<b>b</b>) detected landmarks; (<b>c</b>) detected landmarks that are not on the eyebrows.</p> Full article ">
18 pages, 4920 KiB  
Article
Development of a Non-Contacting Muscular Activity Measurement System for Evaluating Knee Extensors Training in Real-Time
by Zixi Gu, Shengxu Liu, Sarah Cosentino and Atsuo Takanishi
Sensors 2022, 22(12), 4632; https://doi.org/10.3390/s22124632 - 19 Jun 2022
Cited by 4 | Viewed by 2727
Abstract
To give people more specific information on the quality of their daily motion, it is necessary to continuously measure muscular activity during everyday occupations in an easy way. The traditional methods to measure muscle activity using a combination of surface electromyography (sEMG) sensors [...] Read more.
To give people more specific information on the quality of their daily motion, it is necessary to continuously measure muscular activity during everyday occupations in an easy way. The traditional methods to measure muscle activity using a combination of surface electromyography (sEMG) sensors and optical motion capture system are expensive and not suitable for non-technical users and unstructured environment. For this reason, in our group we are researching methods to estimate leg muscle activity using non-contact wearable sensors, improving ease of movement and system usability. In a previous study, we developed a method to estimate muscle activity via only a single inertial measurement unit (IMU) on the shank. In this study, we describe a method to estimate muscle activity during walking via two IMU sensors, using an original sensing system and specifically developed estimation algorithms based on ANN techniques. The muscle activity estimation results, estimated by the proposed algorithm after optimization, showed a relatively high estimation accuracy with a correlation efficient of R2 = 0.48 and a standard deviation STD = 0.10, with a total system average delay of 192 ms. As the average interval between different gait phases in human gait is 250–1000 ms, a 192 ms delay is still acceptable for daily walking requirements. For this reason, compared with the previous study, the newly proposed system presents a higher accuracy and is better suitable for real-time leg muscle activity estimation during walking. Full article
(This article belongs to the Special Issue Recent Advancements in Sensor Technologies for Healthcare and Biomedical Applications)
►▼ Show Figures

Figure 1

Figure 1
<p>The overview of the system.</p> Full article ">Figure 2
<p>WB sensors: (<b>a</b>) WB-EMMG sensor and (<b>b</b>) WB-4 IMU sensor.</p> Full article ">Figure 3
<p>Measurement system using in the experiment.</p> Full article ">Figure 4
<p>The experiment environment.</p> Full article ">Figure 5
<p>MVC measurement method: (<b>a</b>) measurement for the thigh (<b>b</b>) measurement for the calf.</p> Full article ">Figure 6
<p>Position of IMU &amp; sEMG sensors setup in the experiment.</p> Full article ">Figure 7
<p>Overview of muscle activity estimation system.</p> Full article ">Figure 8
<p>The calculation algorithm of gait parameters.</p> Full article ">Figure 9
<p>The motion data example of angular velocity for one foot.</p> Full article ">Figure 10
<p>EKF process flowchart.</p> Full article ">Figure 11
<p>Walking distance of IMU on the ankle.</p> Full article ">Figure 12
<p>Walking distance of IMU on the ankle.</p> Full article ">Figure 13
<p>EMG data processing workflow.</p> Full article ">Figure 14
<p>Dropout method, randomly removing the connection between nodes.</p> Full article ">Figure 15
<p>Model performance boxplot.</p> Full article ">
17 pages, 3651 KiB  
Article
Voltammetric Behaviour of Rhodamine B at a Screen-Printed Carbon Electrode and Its Trace Determination in Environmental Water Samples
by Kevin C. Honeychurch
Sensors 2022, 22(12), 4631; https://doi.org/10.3390/s22124631 - 19 Jun 2022
Cited by 9 | Viewed by 3543
Abstract
The voltammetric behaviour of Rhodamine B was studied at a screen-printed carbon electrode (SPCE), by cyclic and differential pulse voltammetry. Cyclic voltammograms exhibited two reduction peaks (designated R1 and R2) generated from the reduction of the parent compound through, first, one electron reduction [...] Read more.
The voltammetric behaviour of Rhodamine B was studied at a screen-printed carbon electrode (SPCE), by cyclic and differential pulse voltammetry. Cyclic voltammograms exhibited two reduction peaks (designated R1 and R2) generated from the reduction of the parent compound through, first, one electron reduction (R1) to give a radical species, and then a further one-electron, one-proton reduction to give a neutral molecule (R2). On the reverse positive-going scan, two oxidation peaks were observed. The first, O1, resulted from the oxidation of the species generated at R2, and the second, O2, through the one-electron oxidation of the amine group. The nature of the redox reactions was further investigated by observing the effect of scan rate and pH on the voltammetric behaviour. The developed SPCE method was evaluated by carrying out Rhodamine B determinations on a spiked and unspiked environmental water sample. A mean recovery of 94.3% with an associated coefficient of variation of 2.9% was obtained. The performance characteristics indicated that reliable data may be obtained for Rhodamine B measurements in environmental water samples using this approach. Full article
(This article belongs to the Special Issue Screen-Printed Sensors)
►▼ Show Figures

Figure 1

Figure 1
<p>Cyclic voltammogram, obtained at a scan rate of 100 mVs<sup>−1</sup>, for a 1.0 mM solution of Rhodamine B in 0.1 M phosphate at pH 8.3. (<b>a</b>) Starting potential 0.0 V, initial switching potential −1.5 V, second switching potential +1.5 V, final potential −0.5 V and (<b>b</b>) second scan of the same SPCE. Voltammetric conditions as for <a href="#sensors-22-04631-f001" class="html-fig">Figure 1</a>a.</p> Full article ">Figure 2
<p>(<b>a</b>) Starting potential 0.0 V, initial switching potential −1.5 V, second switching potential +0.5 V, final potential −0.5 V and (<b>b</b>) starting potential 0.0 V, initial switching potential −0.7 V, second switching potential +1.5 V, final potential −0.5 V.</p> Full article ">Figure 3
<p>Typical cyclic voltammograms obtained for 1.0 mM Rhodamine B in (<b>a</b>) pH 1.92, (<b>b</b>) 3.51, (<b>c</b>) pH 7.17, (<b>d</b>) pH 8.00, (<b>e</b>) pH 8.78 and (<b>f</b>) pH 11.4.</p> Full article ">Figure 4
<p>Plot of Ep vs. pH for Rhodamine B. Crosses O2; triangles O1; diamonds R1 and squares R2. Voltammetric conditions as <a href="#sensors-22-04631-f001" class="html-fig">Figure 1</a>.</p> Full article ">Figure 5
<p>The pH dependence for the spirolactam form of Rhodamine B.</p> Full article ">Figure 6
<p>Plots of peak current vs. square root of scan rate for (<b>a</b>) O1, (<b>c</b>) O2, (<b>e</b>) R1 and (<b>g</b>) R2. Plots of current function for (<b>b</b>) O1, (<b>d</b>) O2, (<b>f</b>) R1 and (<b>h</b>) R2. Other voltammetric conditions as for <a href="#sensors-22-04631-f001" class="html-fig">Figure 1</a>a.</p> Full article ">Figure 7
<p>Plot of <span class="html-italic">i</span><sub>p</sub> vs. pH for Rhodamine B (<b>a</b>) oxidation processes, O1 and O2 and (<b>b</b>) reduction processes, R1 and R2. Crosses O2; triangles O1; diamonds R1 and squares R2. Voltammetric conditions as <a href="#sensors-22-04631-f001" class="html-fig">Figure 1</a>. Error bars represent plus or minus a standard deviation.</p> Full article ">Figure 8
<p>Proposed scheme for the voltammetric behaviour of Rhodamine B at the SPCE.</p> Full article ">Figure 9
<p>Differential pulse voltammograms obtained with an SPCE for solid line, 2.1 µgmL<sup>−1</sup> Rhodamine B in 0.1 M phosphate buffer pH 7.1. Dashed line, buffer only. Voltammetric conditions: 0.0 V 15 s vs. Ag/AgCl. Pulse repetition time 0.2 s, step height 2.4 mV in the positive potential by applying pulse amplitude of 50 mV, pulse duration 50 ms.</p> Full article ">Figure 10
<p>Effect of Rhodamine B concentration on the differential pulse voltammetric current peak (<span class="html-italic">i</span><sub>p</sub>) for oxidation peak O2. Insert, linear section. Each point is the mean of three separate SPCEs. Error bars represent plus and minus a standard deviation.</p> Full article ">Figure 11
<p>Differential pulse voltammograms peak currents obtained with an SPCE for 2.1 µgmL<sup>−1</sup> Rhodamine B in 0.1 M phosphate buffer pH 7.1 with increasing concentrations of ascorbic acid. Voltammetric conditions as for <a href="#sensors-22-04631-f007" class="html-fig">Figure 7</a>. Each point is the mean of three individual SPCEs. Error bars represent plus and minus a standard deviation.</p> Full article ">Figure 12
<p>Differential pulse voltammogram of Rhodamine B in an environmental water sample adjusted to be 0.1 M pH 7.1 phosphate buffer, 0.8 M ascorbic acid. Insert shows same voltammogram at larger scale.</p> Full article ">Scheme 1
<p>Structure of Rhodamine B.</p> Full article ">
18 pages, 1439 KiB  
Article
Hybrid Technique for Cyber-Physical Security in Cloud-Based Smart Industries
by Deepak Garg, Shalli Rani, Norbert Herencsar, Sahil Verma, Marcin Wozniak and Muhammad Fazal Ijaz
Sensors 2022, 22(12), 4630; https://doi.org/10.3390/s22124630 - 19 Jun 2022
Cited by 14 | Viewed by 2640
Abstract
New technologies and trends in industries have opened up ways for distributed establishment of Cyber-Physical Systems (CPSs) for smart industries. CPSs are largely based upon Internet of Things (IoT) because of data storage on cloud servers which poses many constraints due to the [...] Read more.
New technologies and trends in industries have opened up ways for distributed establishment of Cyber-Physical Systems (CPSs) for smart industries. CPSs are largely based upon Internet of Things (IoT) because of data storage on cloud servers which poses many constraints due to the heterogeneous nature of devices involved in communication. Among other challenges, security is the most daunting challenge that contributes, at least in part, to the impeded momentum of the CPS realization. Designers assume that CPSs are themselves protected as they cannot be accessed from external networks. However, these days, CPSs have combined parts of the cyber world and also the physical layer. Therefore, cyber security problems are large for commercial CPSs because the systems move with one another and conjointly with physical surroundings, i.e., Complex Industrial Applications (CIA). Therefore, in this paper, a novel data security algorithm Dynamic Hybrid Secured Encryption Technique (DHSE) is proposed based on the hybrid encryption scheme of Advanced Encryption Standard (AES), Identity-Based Encryption (IBE) and Attribute-Based Encryption (ABE). The proposed algorithm divides the data into three categories, i.e., less sensitive, mid-sensitive and high sensitive. The data is distributed by forming the named-data packets (NDPs) via labelling the names. One can choose the number of rounds depending on the actual size of a key; it is necessary to perform a minimum of 10 rounds for 128-bit keys in DHSE. The average encryption time taken by AES (Advanced Encryption Standard), IBE (Identity-based encryption) and ABE (Attribute-Based Encryption) is 3.25 ms, 2.18 ms and 2.39 ms, respectively. Whereas the average time taken by the DHSE encryption algorithm is 2.07 ms which is very much less when compared to other algorithms. Similarly, the average decryption times taken by AES, IBE and ABE are 1.77 ms, 1.09 ms and 1.20 ms and the average times taken by the DHSE decryption algorithms are 1.07 ms, which is very much less when compared to other algorithms. The analysis shows that the framework is well designed and provides confidentiality of data with minimum encryption and decryption time. Therefore, the proposed approach is well suited for CPS-IoT. Full article
(This article belongs to the Section Intelligent Sensors)
►▼ Show Figures

Figure 1

Figure 1
<p>Cloud-based architecture of IOT-CPS.</p> Full article ">Figure 2
<p>Working Model of Advanced Encryption Standard (AES).</p> Full article ">Figure 3
<p>Working Model of Identity-Based Encryption (IBE).</p> Full article ">Figure 4
<p>Flow Chart of Attribute-Based Encryption (ABE).</p> Full article ">Figure 5
<p>Document Segregation Process.</p> Full article ">Figure 6
<p>Bitwise encryption time.</p> Full article ">Figure 7
<p>Bitwise decryption time.</p> Full article ">Figure 8
<p>Methodology of TOPSIS.</p> Full article ">Figure 9
<p>Relative importance values.</p> Full article ">Figure 10
<p>Random index (RI) values.</p> Full article ">Figure 11
<p>Relative importance of attributes of proposed approach.</p> Full article ">Figure 12
<p>Ranking of encryption techniques.</p> Full article ">
15 pages, 2870 KiB  
Article
Noise Reduction in Human Motion-Captured Signals for Computer Animation based on B-Spline Filtering
by Mehdi Memar Ardestani and Hong Yan
Sensors 2022, 22(12), 4629; https://doi.org/10.3390/s22124629 - 19 Jun 2022
Cited by 8 | Viewed by 2726
Abstract
Motion capturing is used to record the natural movements of humans for a particular task. The motions recorded are extensively used to produce animation characters with natural movements and for virtual reality (VR) devices. The raw captured motion signals, however, contain noises introduced [...] Read more.
Motion capturing is used to record the natural movements of humans for a particular task. The motions recorded are extensively used to produce animation characters with natural movements and for virtual reality (VR) devices. The raw captured motion signals, however, contain noises introduced during the capturing process. Therefore, the signals should be effectively processed before they can be applied to animation characters. In this study, we analyzed several common methods used for smoothing signals. The smoothed signals were then compared based on the smoothness metrics defined. It was concluded that the filtering based on the B-spline-based least square method could achieve high-quality outputs with predetermined continuity and minimal parameter adjustments for a variety of motion signals. Full article
(This article belongs to the Section Biosensors)
►▼ Show Figures

Figure 1

Figure 1
<p>(<b>a</b>) The experimental true sine signal of frequency 1 and (<b>b</b>) the sine signal with randomly generated noise added.</p> Full article ">Figure 2
<p>The output of the moving average filter with the ground-truth sine signal for comparison using (<b>a</b>) an asymmetric window of size 2, (<b>b</b>) an asymmetric window of size 3, (<b>c</b>) a symmetric window of size 3, and (<b>d</b>) a symmetric window of size 5.</p> Full article ">Figure 3
<p>The output of the B-spline filter with the ground-truth sine signal for comparison using (<b>a</b>) 500, (<b>b</b>) 200, (<b>c</b>) 100, and (<b>d</b>) 50 control points.</p> Full article ">Figure 4
<p>The output of the Kalman filter with the ground-truth sine signal for comparison using (<b>a</b>) <math display="inline"><semantics> <mrow> <msub> <mi>E</mi> <mrow> <mi>m</mi> <mi>e</mi> <mi>a</mi> </mrow> </msub> </mrow> </semantics></math> = 0.01 and (<b>b</b>) <math display="inline"><semantics> <mrow> <msub> <mi>E</mi> <mrow> <mi>m</mi> <mi>e</mi> <mi>a</mi> </mrow> </msub> </mrow> </semantics></math> = 0.02.</p> Full article ">Figure 5
<p>Original noisy motion signal from channel 1 of the root-joint of the kicking character versus the smoothed signal via the moving average filter using (<b>a</b>) an asymmetric window of size 2, (<b>b</b>) an asymmetric window of size 3, (<b>c</b>) a symmetric window of size 3, and (<b>d</b>) a symmetric window of size 5.</p> Full article ">Figure 6
<p>Original noisy motion signal from channel 3 of the root-joint of the kicking character versus the smoothed signal via the moving average filter using (<b>a</b>) an asymmetric window of size 2, (<b>b</b>) an asymmetric window of size 3, (<b>c</b>) a symmetric window of size 3, and (<b>d</b>) a symmetric window of size 5.</p> Full article ">Figure 6 Cont.
<p>Original noisy motion signal from channel 3 of the root-joint of the kicking character versus the smoothed signal via the moving average filter using (<b>a</b>) an asymmetric window of size 2, (<b>b</b>) an asymmetric window of size 3, (<b>c</b>) a symmetric window of size 3, and (<b>d</b>) a symmetric window of size 5.</p> Full article ">Figure 7
<p>Percentages of relative difference for smoothness gained for the selected root-joint signals of the kicking character using the moving average filter with symmetric and asymmetric averaging windows of different sizes.</p> Full article ">Figure 8
<p>Original noisy motion signal from channel 1 of the root-joint of the kicking character versus the smoothed signal via B-spline smoothing using (<b>a</b>) 598, (<b>b</b>) 100, (<b>c</b>) 50, and (<b>d</b>) 25 control points.</p> Full article ">Figure 8 Cont.
<p>Original noisy motion signal from channel 1 of the root-joint of the kicking character versus the smoothed signal via B-spline smoothing using (<b>a</b>) 598, (<b>b</b>) 100, (<b>c</b>) 50, and (<b>d</b>) 25 control points.</p> Full article ">Figure 9
<p>Original noisy motion signal from channel 3 of the root-joint of the kicking character versus the smoothed signal via B-spline smoothing using (<b>a</b>) 598, (<b>b</b>) 100, (<b>c</b>) 50, and (<b>d</b>) 25 control points.</p> Full article ">Figure 10
<p>Original noisy motion signal from channel 1 of the root-joint of the kicking character versus the smoothed signal via the Kalman filter using (<b>a</b>) <math display="inline"><semantics> <mrow> <msub> <mi>E</mi> <mrow> <mi>m</mi> <mi>e</mi> <mi>a</mi> </mrow> </msub> </mrow> </semantics></math> = 0.005, (<b>b</b>) <math display="inline"><semantics> <mrow> <msub> <mi>E</mi> <mrow> <mi>m</mi> <mi>e</mi> <mi>a</mi> </mrow> </msub> </mrow> </semantics></math> = 0.01 and (<b>c</b>) <math display="inline"><semantics> <mrow> <msub> <mi>E</mi> <mrow> <mi>m</mi> <mi>e</mi> <mi>a</mi> </mrow> </msub> </mrow> </semantics></math> = 0.02.</p> Full article ">Figure 11
<p>Original noisy motion signal from channel 3 of the root-joint of the kicking character versus the smoothed signal via the Kalman filter using (<b>a</b>) <math display="inline"><semantics> <mrow> <msub> <mi>E</mi> <mrow> <mi>m</mi> <mi>e</mi> <mi>a</mi> </mrow> </msub> </mrow> </semantics></math> = 0.005, (<b>b</b>) <math display="inline"><semantics> <mrow> <msub> <mi>E</mi> <mrow> <mi>m</mi> <mi>e</mi> <mi>a</mi> </mrow> </msub> </mrow> </semantics></math> = 0.01 and (<b>c</b>) <math display="inline"><semantics> <mrow> <msub> <mi>E</mi> <mrow> <mi>m</mi> <mi>e</mi> <mi>a</mi> </mrow> </msub> </mrow> </semantics></math> = 0.02.</p> Full article ">
22 pages, 49074 KiB  
Article
LiDAR Echo Gaussian Decomposition Algorithm for FPGA Implementation
by Guoqing Zhou, Xiang Zhou, Jinlong Chen, Guoshuai Jia and Qiang Zhu
Sensors 2022, 22(12), 4628; https://doi.org/10.3390/s22124628 - 19 Jun 2022
Cited by 23 | Viewed by 2687
Abstract
As the existing processing algorithms for LiDAR echo decomposition are time-consuming, this paper proposes an FPGA-based improved Gaussian full-waveform decomposition method. The proposed FPGA architecture consists of three modules: (i) a pre-processing module, which is used to pipeline data reading and Gaussian filtering, [...] Read more.
As the existing processing algorithms for LiDAR echo decomposition are time-consuming, this paper proposes an FPGA-based improved Gaussian full-waveform decomposition method. The proposed FPGA architecture consists of three modules: (i) a pre-processing module, which is used to pipeline data reading and Gaussian filtering, (ii) the inflection point coordinate solution module, applied to the second-order differential operation and to calculate inflection point coordinates, and (iii) the Gaussian component parameter solution and echo component positioning module, which is utilized to calculate the Gaussian component and echo time parameters. Finally, two LiDAR datasets, covering the Congo and Antarctic regions, are used to verify the accuracy and speed of the proposed method. The experimental results show that (i) the accuracy of the FPGA-based processing is equivalent to that of PC-based processing, and (ii) the processing speed of the FPGA-based processing is 292 times faster than that of PC-based processing. Full article
(This article belongs to the Section Radar Sensors)
►▼ Show Figures

Figure 1

Figure 1
<p>Comparison of LiDAR waveform and filtered waveform.</p> Full article ">Figure 2
<p>Comparison of the LiDAR waveform and the waveform after the second-order derivation.</p> Full article ">Figure 3
<p>Inflection points near coordinates representation.</p> Full article ">Figure 4
<p>Analog echo waveform.</p> Full article ">Figure 5
<p>FPGA architecture: (<b>a</b>) overall architecture; (<b>b</b>) pre-processing architecture; (<b>c</b>) inflection point coordinate solution architecture; (<b>d</b>) Gaussian component parameter solution and echo component positioning architecture.</p> Full article ">Figure 6
<p>RAM data read module.</p> Full article ">Figure 7
<p>Filter module.</p> Full article ">Figure 8
<p>Second-order differential module.</p> Full article ">Figure 9
<p>Inflection point coordinate query module.</p> Full article ">Figure 10
<p>Schematic diagram of state machine.</p> Full article ">Figure 11
<p>Inflection point coordinate calculation module.</p> Full article ">Figure 12
<p>The calculation for the amplitude of each Gaussian component: (<b>a</b>) Solving for the maximum value, <span class="html-italic">y<sub>max</sub></span>; (<b>b</b>) Solving for the amplitude <math display="inline"><semantics> <mrow> <msub> <mi>a</mi> <mi>i</mi> </msub> </mrow> </semantics></math>.</p> Full article ">Figure 13
<p>Solving center position <math display="inline"><semantics> <mrow> <msub> <mi>c</mi> <mi>i</mi> </msub> </mrow> </semantics></math>, pulse width <math display="inline"><semantics> <mrow> <msub> <mi>δ</mi> <mi>i</mi> </msub> </mrow> </semantics></math>, and result output module.</p> Full article ">Figure 14
<p>LiDAR echo decomposition results in Congo: (<b>a</b>) Central position; (<b>b</b>) Pulse width; (<b>c</b>) Amplitude; (<b>d</b>) Distance measuring point.</p> Full article ">Figure 15
<p>LiDAR echo decomposition results in Antarctica: (<b>a</b>) Central position; (<b>b</b>) Pulse width; (<b>c</b>) Amplitude; (<b>d</b>) Distance measuring point.</p> Full article ">Figure 16
<p>Error analysis of echo waveform.</p> Full article ">Figure A1
<p>(<b>a</b>–<b>t</b>) LiDAR waveform data and LiDAR echo decomposition waveform in the Congo region. The abscissa is the sampling time point of the LiDAR waveform and the ordinate is the amplitude coordinate of the waveform.</p> Full article ">Figure A1 Cont.
<p>(<b>a</b>–<b>t</b>) LiDAR waveform data and LiDAR echo decomposition waveform in the Congo region. The abscissa is the sampling time point of the LiDAR waveform and the ordinate is the amplitude coordinate of the waveform.</p> Full article ">Figure A2
<p>(<b>a</b>–<b>t</b>) LiDAR waveform data and decomposed waveforms of ocean LiDAR waveforms in Antarctica. The abscissa is the sampling time point of the LiDAR waveform and the ordinate is the amplitude coordinate of the waveform.</p> Full article ">Figure A2 Cont.
<p>(<b>a</b>–<b>t</b>) LiDAR waveform data and decomposed waveforms of ocean LiDAR waveforms in Antarctica. The abscissa is the sampling time point of the LiDAR waveform and the ordinate is the amplitude coordinate of the waveform.</p> Full article ">
26 pages, 6924 KiB  
Article
InteliRank: A Four-Pronged Agent for the Intelligent Ranking of Cloud Services Based on End-Users’ Feedback
by Muhammad Munir Ud Din, Nasser Alshammari, Saad Awadh Alanazi, Fahad Ahmad, Shahid Naseem, Muhammad Saleem Khan and Hafiz Syed Imran Haider
Sensors 2022, 22(12), 4627; https://doi.org/10.3390/s22124627 - 19 Jun 2022
Cited by 4 | Viewed by 2782
Abstract
Cloud Computing (CC) provides a combination of technologies that allows the user to use the most resources in the least amount of time and with the least amount of money. CC semantics play a critical role in ranking heterogeneous data by using the [...] Read more.
Cloud Computing (CC) provides a combination of technologies that allows the user to use the most resources in the least amount of time and with the least amount of money. CC semantics play a critical role in ranking heterogeneous data by using the properties of different cloud services and then achieving the optimal cloud service. Regardless of the efforts made to enable simple access to this CC innovation, in the presence of various organizations delivering comparative services at varying cost and execution levels, it is far more difficult to identify the ideal cloud service based on the user’s requirements. In this research, we propose a Cloud-Services-Ranking Agent (CSRA) for analyzing cloud services using end-users’ feedback, including Platform as a Service (PaaS), Infrastructure as a Service (IaaS), and Software as a Service (SaaS), based on ontology mapping and selecting the optimal service. The proposed CSRA possesses Machine-Learning (ML) techniques for ranking cloud services using parameters such as availability, security, reliability, and cost. Here, the Quality of Web Service (QWS) dataset is used, which has seven major cloud services categories, ranked from 0–6, to extract the required persuasive features through Sequential Minimal Optimization Regression (SMOreg). The classification outcomes through SMOreg are capable and demonstrate a general accuracy of around 98.71% in identifying optimum cloud services through the identified parameters. The main advantage of SMOreg is that the amount of memory required for SMO is linear. The findings show that our improved model in terms of precision outperforms prevailing techniques such as Multilayer Perceptron (MLP) and Linear Regression (LR). Full article
(This article belongs to the Special Issue Cloud/Edge/Fog Computing for Network and IoT)
►▼ Show Figures

Figure 1

Figure 1
<p>Cloud based computing and communication.</p> Full article ">Figure 2
<p>Public cloud computing.</p> Full article ">Figure 3
<p>Private cloud computing.</p> Full article ">Figure 4
<p>Community cloud computing.</p> Full article ">Figure 5
<p>InteliRank: Cloud-service ranking agent.</p> Full article ">Figure 6
<p>Cloud-service ranking for identified parameters using Student’s <span class="html-italic">t</span>-test.</p> Full article ">Figure 7
<p>Correlation analyses among identified parameters using end-users’ feedback.</p> Full article ">Figure 8
<p>Parametric distribution of dataset.</p> Full article ">Figure 9
<p>Combined matrix plot.</p> Full article ">Figure 10
<p>(<b>a</b>). Sequential minimal optimization regression error (X: Availability vs. Y: Predicted Ranking). (<b>b</b>). Sequential minimal optimization regression error (X: Security vs. Y: Predicted Ranking). (<b>c</b>). Sequential minimal optimization regression error (X: Reliability vs. Y: Predicted Ranking. (<b>d</b>). Sequential minimal optimization regression error (X: Cost vs. Y: Predicted Ranking). (<b>e</b>). Sequential minimal optimization regression error (X: Ranking vs. Y: Predicted Ranking).</p> Full article ">Figure 10 Cont.
<p>(<b>a</b>). Sequential minimal optimization regression error (X: Availability vs. Y: Predicted Ranking). (<b>b</b>). Sequential minimal optimization regression error (X: Security vs. Y: Predicted Ranking). (<b>c</b>). Sequential minimal optimization regression error (X: Reliability vs. Y: Predicted Ranking. (<b>d</b>). Sequential minimal optimization regression error (X: Cost vs. Y: Predicted Ranking). (<b>e</b>). Sequential minimal optimization regression error (X: Ranking vs. Y: Predicted Ranking).</p> Full article ">Figure 11
<p>(<b>a</b>). Multilayer Perceptron error (X: Availability vs. Y: Predicted Ranking). (<b>b</b>). Multilayer Perceptron error (X: Security vs. Y: Predicted Ranking). (<b>c</b>). Multilayer Perceptron error (X: Reliability vs. Y: Predicted Ranking). (<b>d</b>). Multilayer Perceptron error (X: Cost vs. Y: Predicted Ranking). (<b>e</b>). Multilayer Perceptron error (X: Ranking vs. Y: Predicted Ranking).</p> Full article ">Figure 11 Cont.
<p>(<b>a</b>). Multilayer Perceptron error (X: Availability vs. Y: Predicted Ranking). (<b>b</b>). Multilayer Perceptron error (X: Security vs. Y: Predicted Ranking). (<b>c</b>). Multilayer Perceptron error (X: Reliability vs. Y: Predicted Ranking). (<b>d</b>). Multilayer Perceptron error (X: Cost vs. Y: Predicted Ranking). (<b>e</b>). Multilayer Perceptron error (X: Ranking vs. Y: Predicted Ranking).</p> Full article ">Figure 12
<p>(<b>a</b>). Linear regression error (X: Availability vs. Y: Predicted Ranking). (<b>b</b>). Linear regression error (X: Security vs. Y: Predicted Ranking). (<b>c</b>). Linear regression error (X: Reliability vs. Y: Predicted Ranking). (<b>d</b>). Linear regression error (X: Cost vs. Y: Predicted Ranking). (<b>e</b>). Linear regression error (X: Ranking vs. Y: Predicted Ranking).</p> Full article ">Figure 12 Cont.
<p>(<b>a</b>). Linear regression error (X: Availability vs. Y: Predicted Ranking). (<b>b</b>). Linear regression error (X: Security vs. Y: Predicted Ranking). (<b>c</b>). Linear regression error (X: Reliability vs. Y: Predicted Ranking). (<b>d</b>). Linear regression error (X: Cost vs. Y: Predicted Ranking). (<b>e</b>). Linear regression error (X: Ranking vs. Y: Predicted Ranking).</p> Full article ">Figure 13
<p>(<b>a</b>). Clustering based on availability through self-organizing map. (<b>b</b>). Clustering based on security through self-organizing map. (<b>c</b>). clustering based on reliability through self-organizing map. (<b>d</b>). Clustering based on cost through self-organizing map. (<b>e</b>). Clustering based on ranking through self-organizing map.</p> Full article ">
26 pages, 4484 KiB  
Article
HARNU-Net: Hierarchical Attention Residual Nested U-Net for Change Detection in Remote Sensing Images
by Haojin Li, Liejun Wang and Shuli Cheng
Sensors 2022, 22(12), 4626; https://doi.org/10.3390/s22124626 - 19 Jun 2022
Cited by 6 | Viewed by 2538
Abstract
Change detection (CD) is a particularly important task in the field of remote sensing image processing. It is of practical importance for people when making decisions about transitional situations on the Earth’s surface. The existing CD methods focus on the design of feature [...] Read more.
Change detection (CD) is a particularly important task in the field of remote sensing image processing. It is of practical importance for people when making decisions about transitional situations on the Earth’s surface. The existing CD methods focus on the design of feature extraction network, ignoring the strategy fusion and attention enhancement of the extracted features, which will lead to the problems of incomplete boundary of changed area and missing detection of small targets in the final output change map. To overcome the above problems, we proposed a hierarchical attention residual nested U-Net (HARNU-Net) for remote sensing image CD. First, the backbone network is composed of a Siamese network and nested U-Net. We remold the convolution block in nested U-Net and proposed ACON-Relu residual convolution block (A-R), which reduces the missed detection rate of the backbone network in small change areas. Second, this paper proposed the adjacent feature fusion module (AFFM). Based on the adjacency fusion strategy, the module effectively integrates the details and semantic information of multi-level features, so as to realize the feature complementarity and spatial mutual enhancement between adjacent features. Finally, the hierarchical attention residual module (HARM) is proposed, which locally filters and enhances the features in a more fine-grained space to output a much better change map. Adequate experiments on three challenging benchmark public datasets, CDD, LEVIR-CD and BCDD, show that our method outperforms several other state-of-the-art methods and performs excellent in F1, IOU and visual image quality. Full article
(This article belongs to the Section Sensing and Imaging)
►▼ Show Figures

Figure 1

Figure 1
<p>Architecture of the proposed HARNU-Net. (<b>a</b>) is the backbone of HARNU-Net, used for feature extraction. (<b>b</b>) is an improved convolutional block, proposed to enhance the backbone network performance. AFFM is used for feature fusion, HARM is used for feature filtering and enhancement. Detailed structure of AFFM and HARM are shown in <a href="#sensors-22-04626-f002" class="html-fig">Figure 2</a> and <a href="#sensors-22-04626-f003" class="html-fig">Figure 3</a>.</p> Full article ">Figure 2
<p>Architecture of Adjacent Feature Fusion Module (AFFM).</p> Full article ">Figure 3
<p>Architecture of Hierarchical Attention Residual Module (HARM).</p> Full article ">Figure 4
<p>Architecture of CAM.</p> Full article ">Figure 5
<p>Architecture of SAM.</p> Full article ">Figure 6
<p>Architecture of CBAM.</p> Full article ">Figure 7
<p>Illustration of samples from three dataset. The samples from left to right are in order from CDD, BCDD, LEVIR-CD. T1 and T2 indicate the bi-temporal image pairs. GT indicates the ground truth.</p> Full article ">Figure 8
<p>Visualization results on the CDD dataset. (<b>a</b>) T1 images. (<b>b</b>) T2 images. (<b>c</b>) Ground Truth. (<b>d</b>) FC-EF. (<b>e</b>) FC-Siam-conc. (<b>f</b>) FC-Siam-diff. (<b>g</b>) CDNet. (<b>h</b>) STANet. (<b>i</b>) BIT. (<b>j</b>) SNUNet. (<b>k</b>) Ours. White indicates correctly detected changed areas, black indicates correctly detected unchanged areas, red indicates incorrectly detected unchanged areas as changed areas, and green indicates unpredicted changed areas.</p> Full article ">Figure 9
<p>Visualization results on the BCDD dataset. (<b>a</b>) T1 images. (<b>b</b>) T2 images. (<b>c</b>) Ground Truth. (<b>d</b>) FC-EF. (<b>e</b>) FC-Siam-conc. (<b>f</b>) FC-Siam-diff. (<b>g</b>) CDNet. (<b>h</b>) STANet. (<b>i</b>) BIT. (<b>j</b>) SNUNet. (<b>k</b>) Ours. White indicates correctly detected changed areas, black indicates correctly detected unchanged areas, red indicates incorrectly detected unchanged areas as changed areas, and green indicates unpredicted changed areas.</p> Full article ">Figure 10
<p>Visualization results on the LEVIR-CD dataset. (<b>a</b>) T1 images. (<b>b</b>) T2 images. (<b>c</b>) Ground Truth. (<b>d</b>) FC-EF. (<b>e</b>) FC-Siam-conc. (<b>f</b>) FC-Siam-diff. (<b>g</b>) CDNet. (<b>h</b>) STANet. (<b>i</b>) BIT. (<b>j</b>) SNUNet. (<b>k</b>) Ours. White indicates correctly detected changed areas, black indicates correctly detected unchanged areas, red indicates incorrectly detected unchanged areas as changed areas, and green indicates unpredicted changed areas.</p> Full article ">Figure 11
<p>Visualization results of ablation experiments performed on CDD dataset. (<b>a</b>) T1 images. (<b>b</b>) T2 images. (<b>c</b>) Ground Truth. (<b>d</b>) Baseline. (<b>e</b>) Baseline + A-R. (<b>f</b>) Baseline + AFFM. (<b>g</b>) Baseline + HARM. (<b>h</b>) Baseline + AFFM + HARM. (<b>i</b>) Baseline + A-R + AFFM + HARM. White indicates the predicted change area, black indicates the predicted unchanged area.</p> Full article ">Figure 12
<p>Comparison of attentional heat maps before and after baseline model improvement. (<b>a</b>) T1 image, (<b>b</b>) T2 image, (<b>c</b>) ground truth, (<b>d</b>) attentional heat map generated by the baseline model UNet++, (<b>e</b>) attentional heat map generated by HARNU-Net.</p> Full article ">Figure 13
<p>Validation results F1 of ablation experiments on CDD dataset. Smaller rectangular boxes show a clearer result.</p> Full article ">Figure 14
<p>Validation results F1 for different branches of HARM on the BCDD dataset. Smaller rectangular boxes show a clearer result.</p> Full article ">Figure 15
<p>Ablation experiments performed on the LEVIR-CD dataset applying HARM to other models. All scores are expressed as a percentage (%).</p> Full article ">
Show export options expand_more Show export options expand_less
Select all
Export citation of selected articles as:
 
Displaying articles 1-346
Previous Issue
Next Issue
Back to TopTop