CN118195296A - Scenic spot crowd gathering risk assessment and early warning method and system based on multiple features - Google Patents
Scenic spot crowd gathering risk assessment and early warning method and system based on multiple features Download PDFInfo
- Publication number
- CN118195296A CN118195296A CN202410087133.0A CN202410087133A CN118195296A CN 118195296 A CN118195296 A CN 118195296A CN 202410087133 A CN202410087133 A CN 202410087133A CN 118195296 A CN118195296 A CN 118195296A
- Authority
- CN
- China
- Prior art keywords
- crowd
- scenic spot
- model
- feature
- early warning
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000012502 risk assessment Methods 0.000 title claims abstract description 35
- 238000000034 method Methods 0.000 title claims abstract description 30
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 21
- 238000011156 evaluation Methods 0.000 claims abstract description 16
- 238000012544 monitoring process Methods 0.000 claims abstract description 12
- 230000002776 aggregation Effects 0.000 claims abstract description 11
- 238000004220 aggregation Methods 0.000 claims abstract description 11
- 230000004927 fusion Effects 0.000 claims description 16
- 238000012360 testing method Methods 0.000 claims description 14
- 238000005457 optimization Methods 0.000 claims description 11
- 230000006870 function Effects 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 8
- 238000005065 mining Methods 0.000 claims description 7
- 238000012549 training Methods 0.000 claims description 7
- 201000010099 disease Diseases 0.000 claims description 6
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 5
- 238000010276 construction Methods 0.000 claims description 4
- 238000003860 storage Methods 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 230000000295 complement effect Effects 0.000 claims description 3
- 238000009792 diffusion process Methods 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000001556 precipitation Methods 0.000 claims description 3
- 230000011664 signaling Effects 0.000 claims description 3
- 238000005096 rolling process Methods 0.000 claims description 2
- 238000013135 deep learning Methods 0.000 abstract description 4
- 238000013499 data model Methods 0.000 abstract description 2
- 230000007787 long-term memory Effects 0.000 abstract description 2
- 238000013527 convolutional neural network Methods 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 4
- 230000001932 seasonal effect Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 238000007637 random forest analysis Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000000354 decomposition reaction Methods 0.000 description 2
- 238000013136 deep learning model Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000036541 health Effects 0.000 description 2
- 230000005802 health problem Effects 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 230000015654 memory Effects 0.000 description 1
- YHXISWVBGDMDLQ-UHFFFAOYSA-N moclobemide Chemical compound C1=CC(Cl)=CC=C1C(=O)NCCN1CCOCC1 YHXISWVBGDMDLQ-UHFFFAOYSA-N 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 230000003449 preventive effect Effects 0.000 description 1
- 238000000611 regression analysis Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
- G06V10/449—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters
- G06V10/451—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters with interaction between the filter responses, e.g. cortical complex cells
- G06V10/454—Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
- G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
- G06V20/53—Recognition of crowd images, e.g. recognition of crowd congestion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2123/00—Data types
- G06F2123/02—Data types in the time domain, e.g. time-series data
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Human Resources & Organizations (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Multimedia (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Medical Informatics (AREA)
- Operations Research (AREA)
- Bioinformatics & Computational Biology (AREA)
- Development Economics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Game Theory and Decision Science (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Databases & Information Systems (AREA)
- Marketing (AREA)
- Evolutionary Biology (AREA)
- Quality & Reliability (AREA)
- Biodiversity & Conservation Biology (AREA)
- Human Computer Interaction (AREA)
- Educational Administration (AREA)
Abstract
The invention provides a scenic spot crowd gathering risk assessment and early warning method and system based on multiple characteristics, and relates to the field of model prediction of a deep learning algorithm. The prediction model based on the multi-feature scenic spot crowd risk assessment is a model which is superior to a model constructed by single source data, the multi-feature set comprises a plurality of influencing factors such as passenger flow, weather, other sudden disaster information and the like as feature variable input, the prediction accuracy is effectively improved, and the LSTM long-term memory network algorithm model is used, so that more intelligent passenger flow prediction is effectively realized. In addition, the method combines the historical data with the real-time data, and adds crowd information obtained by real-time monitoring data on the basis of the prediction of the historical data model, so that the prediction model is continuously optimized and updated, the prediction of the evaluation result of the aggregation risk of different time-frequency crowd is realized, and the accuracy and the comprehensive performance of the prediction model in different time periods are greatly improved.
Description
Technical Field
The invention relates to the field of model prediction of a deep learning algorithm, in particular to a scenic spot crowd gathering risk assessment and early warning method and system based on multiple features.
Background
With the increase of economy and the increasing living standard, the method becomes the preferred planning and arrangement of holidays for people. The great potential and the vigorous development trend of the travel industry not only provide different cultural experiences for the masses, but also promote the construction and economic development of local employment, overdraft and infrastructure. However, rapid aggregation of people in a short period of time in a specific area still presents significant risks in managing security, which may include congestion, potential safety hazards, and emergency evacuation difficulties. Therefore, if the crowd gathering risk index in the tourist attraction on the same day can be estimated and predicted in advance according to the data of tourists, weather and the like, the manager of the tourist attraction is assisted to carry out reasonable planning, and corresponding precaution measures are adopted, so that the occurrence of risk accidents can be reduced, and the tourists can enjoy the tourist experience in a safe and orderly environment.
The crowd aggregation risk prediction is to comprehensively consider reasons such as weather and other accident factors on the basis of the prediction of the passenger flow in scenic spots, and to build a model of various risk situations caused by the crowd aggregation possibly occurring at specific time and place by analyzing information such as crowd density, people flow rate and the like. In the state of the art for passenger flow prediction, existing passenger flow prediction models are classified into 3 categories: time series models, economics models, and AI models.
The time series model predicts future traffic conditions by analyzing the time series pattern of historical traffic data. Such models may take into account trend, seasonal, and periodic time characteristics. Common time series models include Auto-REGRESSIVE MOVING AVERAGE MODEL (ARMA), auto-regressive integral moving average (Autoregressive Inte-grated Moving Average Mthod, ARIMA), seasonal decomposition model (Seasonal Decomposition of TIME SERIES, STL), and the like. The empirical model is a predictive model built based on historical data and expert experience. These models are based on past traffic data and domain specific expertise, and simple statistical analysis or rule definition is performed to predict future traffic conditions. Empirical models include averaging, seasonal models, and the like. AI models can be classified into machine learning models and deep learning models. Common machine learning models for passenger flow prediction include regression analysis, support vector machines (Support Vector Regression, SVR), random Forest (RF), and other machine learning, which can better capture complex relationships between data, and are suitable for passenger flow prediction in different scenes. Algorithms such as Long Short-Term Memory (LSTM) and convolutional neural network (Convolutional Neural Networks, CNN) in deep learning are superior to the traditional method in the aspect of passenger flow prediction, but the research quantity and the results based on the technology are still to be enriched.
Crowd counting models are a class of models that utilize deep learning techniques for crowd quantity estimation and counting. The models are mainly applied to image and video data, the number of people is predicted by analyzing human body characteristics in scenes, and the models are suitable for scenes with dense people flow, such as streets, squares, scenic spots and the like. Common crowd counting models include CSRNet(Convolutional Neural Network for Crowd Counting)、MCNN(Multi-column Convolutional Neural Network) and CrowdNet, and the deep learning models have remarkable results in the field of crowd counting and exhibit respective advantages.
Through investigation and analysis, the defects of the current algorithm for predicting the crowd gathering risk of scenic spots are summarized as the following two points:
1. The timeliness of the data is poor: most of the existing models rely on historical passenger flow data to predict, and if the historical data is insufficient or inaccurate, the model prediction will show lower accuracy. The prediction model needs real-time or near real-time data input so as to adjust the prediction result in time and increase the credibility and comprehensive performance of the model result.
2. Analysis of lack of multivariate information: the crowd gathering risk not only needs to consider the congestion degree caused by the crowd density, the crowd speed and the like, but also can change under the condition of considering different external objective factors (weather, traffic, diseases and the like).
Disclosure of Invention
The invention aims to: the method comprises the steps of providing a scenic spot crowd gathering risk assessment and early warning method based on multiple characteristics, and further providing a system for realizing the method, wherein the model predicts scenic spot daily passenger flow according to historical passenger flow data and holiday travel data; then, weather features and other accident features are fused, and the crowd gathering risk in the scenic spot is evaluated and early-warned; finally, according to the analysis of the real-time monitoring image data on the same day, the model is corrected and optimized in time, so that the prediction accuracy is improved, and the problems in the prior art are solved.
In a first aspect, a scenic spot crowd gathering risk assessment and early warning method based on multiple features is provided, including the following steps:
s1, acquiring passenger flow crowd characteristics, weather characteristics and other emergency characteristics;
S2, splicing and fusing the passenger flow crowd characteristics, the weather characteristics and other emergency characteristics to form a multi-feature set;
s3, processing the multi-element feature set based on the LSTM grid unit, generating hidden state output corresponding to the time sequence, mining depth hidden features, and outputting an initial prediction model;
s4, acquiring a real-time crowd image through scenic spot monitoring, preprocessing the real-time crowd image, generating a crowd density estimation graph, and obtaining real-time crowd data characteristics;
s5, adding real-time crowd data features into the initial prediction model, combining the real-time crowd data features with the multi-feature set, and completing model optimization by updating parameters of a crowd aggregation risk assessment model to obtain an optimized prediction model;
S6, predicting new data with labels by using the optimized prediction model to obtain a test risk evaluation coefficient;
and S7, dividing early warning grades according to the test risk evaluation coefficient and outputting the early warning grades.
In a further embodiment of the first aspect, the process of obtaining the crowd feature of the passenger flow in step S1 includes:
starting from the range of the urban area, calculating the arrival rate D i of each national urban area to the scenic spot of the test point by utilizing the mobile phone signaling data and combining the map data:
Di＝Yi/Zi
Wherein Y i is the number of the passenger flows of the telecommunication users in a certain city to reach the trial city, and Z i is the total number of the telecommunication users in the certain city;
Grouping the arrival rates of all scenic spots, namely grouping D Group of j, measuring and calculating the comprehensive arrival rate M Group of j of each group, and measuring and calculating the total amount M Total (S) of national tourists arriving on the same day as the scenic spot reception:
M Group of j＝∑D Group of j*Ri
M Total (S) ＝∑M Group of j
Wherein R i represents the total number of people in a certain city i;
the formula is as follows, in combination with the number of local overnight guests:
M＝Mi*r+Mi-1
Wherein M i represents the amount of tourists arriving on the day of reception; r is the tourist overnight rate; m i-1 is the overnight tourist volume left in the last day, namely the total daily reception tourist volume is the total daily arrival tourist volume and the tourist volume in the scenic spot is further measured and calculated in the last day; and acquiring the relevant crowd characteristics.
In a further embodiment of the first aspect, the acquiring of the weather feature in step S1 includes:
and (3) arranging historical weather data of the preset scenic spot to form a characteristic set, wherein the characteristic set comprises temperature, humidity, air pressure, wind speed, wind direction, visibility and precipitation.
In a further embodiment of the first aspect, the acquiring of the other incident characteristics in step S1 includes:
and acquiring historical geological disaster, traffic and disease information of the city where the predetermined scenic spot is located, integrating and summarizing the historical geological disaster, traffic and disease information into other emergency feature sets, wherein the feature sets comprise occurrence time, place information, whether diffusion or not and influence range dimension.
In a further embodiment of the first aspect, step S2 further comprises:
Splicing and fusing the passenger flow crowd characteristics, the weather characteristics and other emergency characteristics to form a multi-feature set, wherein the expression of the multi-feature set is as follows:
Fusion＝[W1F1,W2F2,W3F3]
Wherein F 1,F2,F3 respectively represents passenger flow crowd, weather and other sudden event feature sets; w 1,W2,W3 represents the weight ratio of each feature;
Thereby forming a new feature vector Fusion, the dimension of which is n 1+n2+n3;
And taking the multi-feature set as a historical data training sample of the crowd gathering risk assessment early warning model.
In a further embodiment of the first aspect, step S3 further includes:
processing the feature vector Fusion based on the LSTM grid unit to generate hidden state output corresponding to the time sequence, and further mining depth hidden features, namely completing hidden layer work of an algorithm model;
And using the full connection layer and applying a linear activation function to process, converting the multidimensional output vector into a one-dimensional vector, and outputting an initial prediction model.
In a further embodiment of the first aspect, step S4 further includes:
Acquiring a real-time crowd image through scenic spot monitoring, denoising and normalizing the real-time crowd image to obtain an input image;
Population count the input images: firstly, extracting head features of a person through a rolling and pooling layer, wherein the features capture different information features in an input image on different scales; then, carrying out weighted average fusion on the extracted characteristic images of each column to form a complementary characteristic trend; finally, the fused feature map is used for further generating a crowd density estimation map, and then the crowd density estimation map is compared with the label density map to calculate loss, and a loss function is defined as follows:
Wherein Θ is a learning parameter in MCNN algorithm, N is the number of training images, X i is an input image, F i is an actual crowd density image, F (X i; Θ) is an estimated density image generated by the algorithm, so as to calculate an L (Θ) loss function for optimizing and updating network parameters;
integrating the output crowd density estimation graph to obtain the total crowd in the graph, wherein the calculation formula is as follows;
crowd total = ≡ DF(Xi, Θ) dxdy.
In a further embodiment of the first aspect, step S7 further includes:
Determining threshold setting of risk coefficients, and dividing early warning levels into four levels which are low risk, medium risk, high risk and extremely high risk according to the threshold;
And (5) comparing the division rules, and converting the risk evaluation coefficient into a corresponding early warning grade.
The explanation of the risk classification correspondence is as follows:
low risk: the situation representing the crowd gathering is relatively safe, and the probability of possible risk problems is low. At low risk levels, crowd-intensive areas may not have obvious crowding, safety hazards, health concerns, etc.
Risk of (1): the situation representing crowd gathering may present some potential risks, but is not very serious. At the risk level, congestion, safety hazards, health problems, etc. may occur, and some measures need to be taken to manage and deal with.
High risk: situations representing crowd gathering may present serious risks, requiring high attention and emergency handling. Under high risk level, serious problems such as congestion, trampling, traffic jam, potential safety hazard and the like can occur, and measures need to be immediately taken to avoid accidents.
Extremely high risk: representing a situation of crowd gathering presents a great risk, possibly leading to serious accidents and problems. Under extremely high risk level, dangerous situations such as emergency evacuation difficulty, casualties and the like can exist, and emergency measures need to be immediately taken to ensure personnel safety.
In a second aspect of the present invention, a scenic spot crowd gathering risk assessment and early warning system is provided, the early warning system includes:
the feature acquisition module is used for acquiring the crowd feature, weather feature and other emergency features of the passenger flow;
The feature fusion module is used for splicing and fusing the passenger flow crowd features, the weather features and other emergency features to form a multi-feature set;
The model construction module is used for processing the multi-element feature set based on the LSTM grid unit, generating hidden state output corresponding to the time sequence, mining depth hidden features and outputting an initial prediction model;
The model optimization module is used for acquiring real-time crowd images through scenic spot monitoring, adding real-time crowd data features into the initial prediction model, combining the real-time crowd data features with the multi-feature set, and completing model optimization by updating parameters of the crowd aggregation risk assessment model to obtain an optimized prediction model;
The prediction module is used for predicting new data with labels by using the optimized prediction model to obtain a test risk evaluation coefficient;
and the early warning output module is used for dividing early warning grades according to the test risk evaluation coefficient and outputting the grade.
In a third aspect of the present invention, a computer readable storage medium is provided, where at least one executable instruction is stored, where the executable instruction when executed on an electronic device causes the electronic device to perform the scenic spot crowd gathering risk assessment and warning method according to the first aspect.
Compared with the prior art, the invention has the following beneficial effects:
1. crowd counting is combined with timing. Related information of crowd counting is added when the time sequence problem is processed, and the predicted information obtained by MCNN through the spatial characteristics of the crowd images is added to an LSTM algorithm for time sequence modeling. The combination of the two algorithms is favorable for analyzing the change trend of people in different time periods and predicting the number of people in the future, and is hopeful to improve the comprehensive performance of the prediction model.
2. The model based on the multi-element feature scenic spot crowd risk assessment prediction is a model superior to a model constructed by single source data, the multi-element feature set comprises a plurality of influencing factors such as passenger flow, weather, other sudden disaster information and the like as feature variable input, the accuracy of prediction is effectively improved, and the LSTM long-term memory network algorithm model is used, so that more intelligent passenger flow prediction is effectively realized.
3. The historical data and the real-time data are combined, and the crowd information obtained by the real-time monitoring data is added on the basis of the prediction of the historical data model, so that the prediction model is continuously optimized and updated, the prediction of the evaluation results of the crowd aggregation risks with different time frequencies is realized, and the accuracy and the comprehensive performance of the prediction model in different time periods are greatly improved.
Drawings
FIG. 1 is a schematic flow chart of a risk assessment prediction model according to an embodiment of the present invention.
Fig. 2 is a schematic diagram of a scenic spot crowd gathering risk assessment and early warning system according to an embodiment of the invention.
Detailed Description
In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the present invention. It will be apparent, however, to one skilled in the art that the invention may be practiced without one or more of these details. In other instances, well-known features have not been described in detail in order to avoid obscuring the invention.
In order to solve the defects of the existing algorithm for predicting the crowd gathering risk of scenic spots, the embodiment discloses a scenic spot crowd gathering risk assessment and early warning method based on multiple characteristics, firstly, the model predicts the daily passenger flow of the scenic spot according to historical passenger flow data and holiday travel data; then, weather features and other accident features are fused, and the crowd gathering risk in the scenic spot is evaluated and early-warned; and finally, carrying out timely correction and optimization on the model according to the analysis of the real-time monitoring image data on the same day, thereby improving the prediction accuracy. The main purpose can be divided into two points:
1. Combining the historical data with the real-time data to form feature complementation, continuously correcting and optimizing a prediction model, and improving the accuracy of the scenic spot crowd gathering risk assessment and early warning and the comprehensive performance of the model;
2. The scenic spot manager is assisted to recognize and cope with potential risks in an early stage, preventive measures are deployed in time, dangerous accidents are reduced, and therefore public safety and tourist management efficiency are improved.
A flow diagram of a scenic spot crowd gathering risk assessment early warning model based on multiple characteristics is shown in fig. 1, and the implementation steps of the specific technology are as follows:
step1: and obtaining the crowd characteristics of the passenger flow. Starting from the range of the urban area, calculating the arrival rate of each national urban area to the scenic spot of the test point by utilizing mobile phone signaling data and combining multi-source heterogeneous data such as map data, wherein the calculation formula is as follows:
Di＝Yi/Zi
Wherein D i is the arrival rate of tourists in a city reaching the trial-point city, Y i is the number of the tourists in the city where the telecommunication user reaches the trial-point city, and Z i is the total number of the telecommunication users in the city.
And grouping the arrival rates of all the scenic spots by using a K_means unsupervised box model, and measuring and calculating the comprehensive arrival rate of each group by means of a K_means clustering weighting algorithm, so as to measure and calculate the total amount of national tourists arriving on the same day of scenic spot reception. The calculation formula is as follows:
M Group of j＝∑D Group of j*Ri
M Total (S) ＝∑M Group of j
Where R i represents the total number of people in a particular city i, the number of people arriving M Total (S) is calculated.
The formula is as follows, in combination with the number of local overnight guests:
M＝Mi*r+Mi-1
Wherein, M i receives the amount of tourists arriving on the same day, r is the rate of overnight tourists, M i-1 is the amount of overnight tourists left on the same day, namely the total amount of tourists received on the same day is the total amount of tourists arriving on the same day and the amount of tourists arriving on the same day are further measured and calculated on the same day, and the relevant crowd characteristics are obtained;
and acquiring weather characteristics. The historical weather data of the scenic spot is arranged to form a characteristic set, such as temperature, humidity, air pressure, wind speed, wind direction, visibility, precipitation and the like;
other incident characteristics are acquired. Information such as geological disasters, traffic, diseases and the like is integrated and summarized into other emergency feature sets, wherein the feature sets comprise dimensions such as occurrence time, place information, whether diffusion, influence range and the like;
Step2: and (5) feature fusion and splicing. The three main types of characteristics (passenger flow crowd, weather and other sudden event characteristics) respectively acquired in Step1 are spliced and fused to form a multi-element characteristic set, and the expression is as follows:
Fusion＝[W1F1,W2F2,W3F3]
Wherein F 1,F2,F3 represents the passenger flow crowd, weather and other sudden event feature sets, and W 1,W2,W3 represents the weight ratio of each feature. This results in a new feature vector Fusion, with dimension n 1+n2+n3. The multi-feature set is used as a historical data training sample of a crowd gathering risk assessment early warning model;
step3: model training, processing the feature set input in the steps based on the LSTM grid unit, generating hidden state output corresponding to the time sequence after processing, and further mining depth hidden features, namely completing hidden layer work of the algorithm model. Then using a full connection layer and applying a linear activation function to process, and converting the multidimensional output vector into a one-dimensional vector, namely finishing the output of the algorithm model;
Step4: real-time crowd data is added. The real-time crowd image is obtained through scenic spot monitoring, and preprocessing is carried out on the image, including denoising, normalization and the like. The population count is performed on the input pictures using MCNN algorithm. First, the features of the human head are extracted by the convolution and pooling layer, which captures different information features in the image at different scales. And then carrying out weighted average fusion on the extracted characteristic images of each column to form complementary characteristic trends. Finally, the fused feature map is used for further generating a crowd density estimation map, and then the crowd density estimation map is compared with the label density map to calculate loss, and a loss function is defined as follows:
Wherein Θ is a learning parameter in MCNN algorithm, N is the number of training images, X i is an input image, F i is an actual crowd density image, F (X i; Θ) is an estimated density image generated by the algorithm, so as to calculate an L (Θ) loss function for optimizing and updating network parameters. Integrating the output crowd density estimation graph to obtain the total crowd in the graph, wherein the calculation formula is as follows;
Step5: and optimizing the prediction model. Combining the acquired real-time crowd data features with the historical data features, sending the combined real-time crowd data features into the linear classifier again, and completing model optimization by updating parameters of the crowd aggregation risk assessment model;
Step6: and obtaining output layer data. Predicting new data with labels by using the trained model to obtain a test risk evaluation coefficient;
Step7: and (5) early warning grade division. And determining threshold setting of risk coefficients, and dividing the early warning level into four levels according to the threshold, wherein the low risk, the medium risk, the high risk and the extremely high risk are respectively identified, and specific dividing rules are shown in table 1. Comparing the division rules, and converting the risk evaluation coefficients into corresponding early warning grades;
TABLE 1 early warning level partitioning based on risk probability coefficients
The explanation of the risk classification correspondence is as follows:
low risk: the situation representing the crowd gathering is relatively safe, and the probability of possible risk problems is low. At low risk levels, crowd-intensive areas may not have obvious crowding, safety hazards, health concerns, etc.
Risk of (1): the situation representing crowd gathering may present some potential risks, but is not very serious. At the risk level, congestion, safety hazards, health problems, etc. may occur, and some measures need to be taken to manage and deal with.
High risk: situations representing crowd gathering may present serious risks, requiring high attention and emergency handling. Under high risk level, serious problems such as congestion, trampling, traffic jam, potential safety hazard and the like can occur, and measures need to be immediately taken to avoid accidents.
Extremely high risk: representing a situation of crowd gathering presents a great risk, possibly leading to serious accidents and problems. Under extremely high risk level, dangerous situations such as emergency evacuation difficulty, casualties and the like can exist, and emergency measures need to be immediately taken to ensure personnel safety.
As a preferred embodiment, a scenic spot crowd gathering risk assessment and early warning system 800 is proposed, see fig. 2, which includes a feature acquisition module 801, a feature fusion module 802, a model construction module 803, a model optimization module 804, a prediction module 805, and an early warning output module 806. The feature acquisition module 801 is configured to acquire crowd features of passenger flow, weather features, and other emergency features. The feature fusion module 802 is configured to splice and fuse the crowd feature, weather feature, and other emergency features of the passenger flow to form a multi-feature set. The model building module 803 is configured to process the multiple feature sets based on the LSTM grid unit, generate a hidden state output corresponding to the time sequence, mine out depth hidden features, and output an initial prediction model. The model optimization module 804 is configured to obtain a real-time crowd image through scenic spot monitoring, add real-time crowd data features to the initial prediction model, combine the real-time crowd data features with the multiple feature sets, and complete model optimization by updating parameters of the crowd aggregation risk assessment model, so as to obtain an optimized prediction model. The prediction module 805 is configured to predict new tagged data using the optimized prediction model to obtain a test risk evaluation coefficient. The early warning output module 806 is configured to divide early warning levels and output the early warning levels according to the test risk evaluation coefficient.
As a preferred embodiment, a computer readable storage medium is provided, where at least one executable instruction is stored, where the executable instruction when executed on an electronic device causes the electronic device to perform the scenic spot crowd gathering risk assessment and warning method based on the multiple features as described in the foregoing embodiment.
The functional blocks shown in the above-described structural block diagrams may be implemented in hardware, software, firmware, or a combination thereof. When implemented in hardware, it may be, for example, an electronic circuit, an Application Specific Integrated Circuit (ASIC), suitable firmware, a plug-in, a function card, or the like. When implemented in software, the elements of the application are the programs or code segments used to perform the required tasks. The program or code segments may be stored in a machine readable medium or transmitted over transmission media or communication links by a data signal carried in a carrier wave. A "machine-readable medium" may include any medium that can store or transfer information. Examples of machine-readable media include electronic circuitry, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio Frequency (RF) links, and the like. The code segments may be downloaded via computer networks such as the internet, intranets, etc.
As described above, although the present invention has been shown and described with reference to certain preferred embodiments, it is not to be construed as limiting the invention itself. Various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (10)
1. The scenic spot crowd gathering risk assessment and early warning method based on the multiple features is characterized by comprising the following steps:
s1, acquiring passenger flow crowd characteristics, weather characteristics and other emergency characteristics;
S2, splicing and fusing the passenger flow crowd characteristics, the weather characteristics and other emergency characteristics to form a multi-feature set;
s3, processing the multi-element feature set based on the LSTM grid unit, generating hidden state output corresponding to the time sequence, mining depth hidden features, and outputting an initial prediction model;
s4, acquiring a real-time crowd image through scenic spot monitoring, preprocessing the real-time crowd image, generating a crowd density estimation graph, and obtaining real-time crowd data characteristics;
s5, adding real-time crowd data features into the initial prediction model, combining the real-time crowd data features with the multi-feature set, and completing model optimization by updating parameters of a crowd aggregation risk assessment model to obtain an optimized prediction model;
S6, predicting new data with labels by using the optimized prediction model to obtain a test risk evaluation coefficient;
and S7, dividing early warning grades according to the test risk evaluation coefficient and outputting the early warning grades.
2. The scenic spot crowd gathering risk assessment and early warning method according to claim 1, wherein the process of acquiring the passenger flow crowd characteristics in step S1 includes:
starting from the range of the urban area, calculating the arrival rate D i of each national urban area to the scenic spot of the test point by utilizing the mobile phone signaling data and combining the map data:
D1＝Y1/Z1
Wherein Y i is the number of the passenger flows of the telecommunication users in a certain city to reach the trial city, and Z i is the total number of the telecommunication users in the certain city;
Grouping the arrival rates of all scenic spots, namely grouping D Group of j, measuring and calculating the comprehensive arrival rate M Group of j of each group, and measuring and calculating the total amount M Total (S) of national tourists arriving on the same day as the scenic spot reception:
M Group of i＝∑D Group of j*Ri
M Total (S) ＝∑M Group of j
Wherein R i represents the total number of people in a certain city i;
the formula is as follows, in combination with the number of local overnight guests:
M＝Mi*r+Mi-1
Wherein M i represents the amount of tourists arriving on the day of reception; r is the tourist overnight rate; m i-1 is the overnight tourist volume left in the last day, namely the total daily reception tourist volume is the total daily arrival tourist volume and the tourist volume in the scenic spot is further measured and calculated in the last day; and acquiring the relevant crowd characteristics.
3. The scenic spot crowd gathering risk assessment and warning method according to claim 1, wherein the acquiring of the weather feature in step S1 includes:
and (3) arranging historical weather data of the preset scenic spot to form a characteristic set, wherein the characteristic set comprises temperature, humidity, air pressure, wind speed, wind direction, visibility and precipitation.
4. The scenic spot crowd gathering risk assessment and early warning method according to claim 1, wherein the acquiring of the other emergency features in step S1 includes:
and acquiring historical geological disaster, traffic and disease information of the city where the predetermined scenic spot is located, integrating and summarizing the historical geological disaster, traffic and disease information into other emergency feature sets, wherein the feature sets comprise occurrence time, place information, whether diffusion or not and influence range dimension.
5. The scenic spot group risk assessment and early warning method according to claim 1, wherein step S2 further comprises:
Splicing and fusing the passenger flow crowd characteristics, the weather characteristics and other emergency characteristics to form a multi-feature set, wherein the expression of the multi-feature set is as follows:
Fusion＝[W1F1,W2F2,W3F3]
Wherein F 1,F2,F3 respectively represents passenger flow crowd, weather and other sudden event feature sets; w 1,W2,W3 represents the weight ratio of each feature;
Thereby forming a new feature vector Fusion, the dimension of which is n 1+n2+n3;
And taking the multi-feature set as a historical data training sample of the crowd gathering risk assessment early warning model.
6. The method of claim 5, wherein step S3 further comprises:
processing the feature vector Fusion based on the LSTM grid unit to generate hidden state output corresponding to the time sequence, and further mining depth hidden features, namely completing hidden layer work of an algorithm model;
And using the full connection layer and applying a linear activation function to process, converting the multidimensional output vector into a one-dimensional vector, and outputting an initial prediction model.
7. The scenic spot group risk assessment and early warning method according to claim 1, wherein step S4 further comprises:
Acquiring a real-time crowd image through scenic spot monitoring, denoising and normalizing the real-time crowd image to obtain an input image;
Population count the input images: firstly, extracting head features of a person through a rolling and pooling layer, wherein the features capture different information features in an input image on different scales; then, carrying out weighted average fusion on the extracted characteristic images of each column to form a complementary characteristic trend; finally, the fused feature map is used for further generating a crowd density estimation map, and then the crowd density estimation map is compared with the label density map to calculate loss, and a loss function is defined as follows:
Wherein Θ is a learning parameter in MCNN algorithm, N is the number of training images, X i is an input image, F i is an actual crowd density image, F (X i; Θ) is an estimated density image generated by the algorithm, so as to calculate an L (Θ) loss function for optimizing and updating network parameters;
integrating the output crowd density estimation graph to obtain the total crowd in the graph, wherein the calculation formula is as follows;
crowd total = ≡ DF(Xi, Θ) dxdy.
8. The scenic spot group risk assessment and early warning method according to claim 1, wherein step S7 further comprises:
Determining threshold setting of risk coefficients, and dividing early warning levels into four levels which are low risk, medium risk, high risk and extremely high risk according to the threshold;
And (5) comparing the division rules, and converting the risk evaluation coefficient into a corresponding early warning grade.
9. The utility model provides a scenic spot crowd gathers risk assessment early warning system which characterized in that includes:
the feature acquisition module is used for acquiring the crowd feature, weather feature and other emergency features of the passenger flow;
The feature fusion module is used for splicing and fusing the passenger flow crowd features, the weather features and other emergency features to form a multi-feature set;
The model construction module is used for processing the multi-element feature set based on the LSTM grid unit, generating hidden state output corresponding to the time sequence, mining depth hidden features and outputting an initial prediction model;
The model optimization module is used for acquiring real-time crowd images through scenic spot monitoring, adding real-time crowd data features into the initial prediction model, combining the real-time crowd data features with the multi-feature set, and completing model optimization by updating parameters of the crowd aggregation risk assessment model to obtain an optimized prediction model;
The prediction module is used for predicting new data with labels by using the optimized prediction model to obtain a test risk evaluation coefficient;
and the early warning output module is used for dividing early warning grades according to the test risk evaluation coefficient and outputting the grade.
10. A computer readable storage medium, wherein at least one executable instruction is stored in the storage medium, which when executed on an electronic device, causes the electronic device to perform the scenic spot crowd gathering risk assessment warning method as claimed in any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410087133.0A CN118195296A (en) | 2024-01-22 | 2024-01-22 | Scenic spot crowd gathering risk assessment and early warning method and system based on multiple features |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410087133.0A CN118195296A (en) | 2024-01-22 | 2024-01-22 | Scenic spot crowd gathering risk assessment and early warning method and system based on multiple features |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN118195296A true CN118195296A (en) | 2024-06-14 |
Family
ID=91402581
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410087133.0A Pending CN118195296A (en) | 2024-01-22 | 2024-01-22 | Scenic spot crowd gathering risk assessment and early warning method and system based on multiple features |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118195296A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118822283A (en) * | 2024-09-19 | 2024-10-22 | 云途信息科技(杭州)有限公司 | Dynamic assessment method and system for road subsidence risk based on multi-factors integrated with pavement representation |
-
2024
- 2024-01-22 CN CN202410087133.0A patent/CN118195296A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118822283A (en) * | 2024-09-19 | 2024-10-22 | 云途信息科技(杭州)有限公司 | Dynamic assessment method and system for road subsidence risk based on multi-factors integrated with pavement representation |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107045788B (en) | Traffic road condition prediction method and device | |
| CN105761488B (en) | Real-time extreme learning machine Short-time Traffic Flow Forecasting Methods based on fusion | |
| CN111582559B (en) | Arrival time estimation method and device | |
| CN116504076A (en) | Expressway traffic flow prediction method based on ETC portal data | |
| CN115953186B (en) | Network appointment vehicle demand pattern recognition and short-time demand prediction method | |
| CN106507315A (en) | A method and system for predicting urban traffic accidents based on network social media data | |
| CN116524712A (en) | Expressway Congestion Prediction Method, System and Device Fusion of Spatiotemporal Correlation Data | |
| CN117558124A (en) | Channel running state deduction prediction method based on multi-source information depth fusion | |
| CN117079462B (en) | Road section emergency traffic event prediction system and method based on Apriori algorithm | |
| CN118115427B (en) | A road surface status detection method and system based on multi-source data | |
| CN114596709B (en) | Data processing method, device, equipment and storage medium | |
| CN116992305A (en) | Weather forecast method and system based on big data | |
| CN116631186A (en) | Highway traffic accident risk assessment method and system based on dangerous driving event data | |
| CN118195296A (en) | Scenic spot crowd gathering risk assessment and early warning method and system based on multiple features | |
| CN117787726A (en) | A road natural disaster risk assessment method based on supervised learning | |
| Saha et al. | Deep learning approach for predictive analytics to support diversion during freeway incidents | |
| KR102359902B1 (en) | Crossroads LOS Prediction Method Based on Big Data and AI, and Storage Medium Having the Same | |
| Thu et al. | Multi-source data analysis for bike sharing systems | |
| CN114139984A (en) | Risk prediction method of urban traffic accident based on collaborative perception of traffic flow and accident | |
| CN117132958B (en) | Road element identification method and related device | |
| CN117689210A (en) | Intelligent Assessment Method for Road Natural Disaster Risks | |
| CN116681278A (en) | Meteorological disaster travel risk assessment method and application thereof | |
| Ma et al. | Research on highway traffic flow prediction based on a hybrid model of ARIMA–GWO–LSTM | |
| CN115619038A (en) | Data-driven highway network congestion prediction and cause analysis method and system | |
| Turki et al. | Using a new algorithm in Machine learning Approaches to estimate level-of-service in hourly traffic flow data in vehicular ad hoc networks |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |