[go: up one dir, main page]

CN111626203B - Railway foreign matter identification method and system based on machine learning - Google Patents

Railway foreign matter identification method and system based on machine learning Download PDF

Info

Publication number
CN111626203B
CN111626203B CN202010460331.9A CN202010460331A CN111626203B CN 111626203 B CN111626203 B CN 111626203B CN 202010460331 A CN202010460331 A CN 202010460331A CN 111626203 B CN111626203 B CN 111626203B
Authority
CN
China
Prior art keywords
image
foreign matter
dangerous area
pixel
layer
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.)
Active
Application number
CN202010460331.9A
Other languages
Chinese (zh)
Other versions
CN111626203A (en
Inventor
苑贵全
骞一凡
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhangjiakou Dongchu Technology Co ltd
Original Assignee
Beijing Weijie Dongbo Information Technology Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Beijing Weijie Dongbo Information Technology Co ltd filed Critical Beijing Weijie Dongbo Information Technology Co ltd
Priority to CN202010460331.9A priority Critical patent/CN111626203B/en
Priority to ZA2020/04403A priority patent/ZA202004403B/en
Publication of CN111626203A publication Critical patent/CN111626203A/en
Application granted granted Critical
Publication of CN111626203B publication Critical patent/CN111626203B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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
    • G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
    • G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B61—RAILWAYS
    • B61L—GUIDING RAILWAY TRAFFIC; ENSURING THE SAFETY OF RAILWAY TRAFFIC
    • B61L23/00—Control, warning or like safety means along the route or between vehicles or trains
    • B61L23/04—Control, warning or like safety means along the route or between vehicles or trains for monitoring the mechanical state of the route
    • B61L23/041—Obstacle detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • 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
    • G06N3/045—Combinations of networks
    • 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
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
    • G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/30—Noise filtering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/34—Smoothing or thinning of the pattern; Morphological operations; Skeletonisation
    • G—PHYSICS
    • G08—SIGNALLING
    • G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B13/00—Burglar, theft or intruder alarms
    • G08B13/18—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength
    • G08B13/189—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems
    • G08B13/194—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems
    • G08B13/196—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems using television cameras
    • G08B13/19602—Image analysis to detect motion of the intruder, e.g. by frame subtraction
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N7/00—Television systems
    • H04N7/18—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Multimedia (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Molecular Biology (AREA)
  • Software Systems (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Mechanical Engineering (AREA)
  • Signal Processing (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)

Abstract

The application provides a railway foreign matter identification method and a system based on machine learning, and the method comprises the following steps: acquiring a monitoring video of a track in front of a train in the running process of the train, and acquiring a plurality of frames of monitoring images in the monitoring video in real time; establishing a dangerous area foreign matter detection window in the monitoring image, and extracting a track dangerous area detection image of the monitoring image in the dangerous area foreign matter detection window; preprocessing a track dangerous area detection image; inputting the preprocessed rail dangerous area detection image into a pre-established foreign matter identification classification model, identifying whether foreign matters exist in the rail dangerous area detection image or not by the foreign matter identification classification model, sending an alarm signal if the foreign matters exist in the rail dangerous area detection image, and otherwise, continuously identifying other rail dangerous area detection images. The detection accuracy of this application to the railway foreign matter is high, uses manpower sparingly, and automatic alarm, the cost is lower, and receives environmental impact little.

Description

Railway foreign matter identification method and system based on machine learning
Technical Field
The application relates to the technical field of railway foreign matter intrusion monitoring, in particular to a railway foreign matter identification method and system based on machine learning.
Background
The distribution area of the high-speed railway in China is wide, the distance is long, the geological condition is complex, the natural disaster is serious, and meanwhile, the highway has the characteristics of day and night operation, closed type, only allowing trains to run and the like. Due to the characteristics of long distance, wide distribution area, day and night operation, closed type and the like, when natural disasters such as landslide, geological subsidence and the like occur along the high-speed railway, road traffic fault warning cannot be issued in time, and serious traffic safety accidents and chain reactions are easily caused. In order to reduce traffic accidents of high-speed railways and reduce casualties and property loss, natural disaster early warning and special traffic safety accident alarming along the high-speed railways are urgently needed to be solved.
Rock rolling, foreign matter such as pedestrian or animal invade the railway boundary, have proruption nature, irregular and unpredictable nature, and the railway traffic accident is aroused frequently, has threatened people's life safety seriously, and traditional track detection mainly relies on the manpower, comes the invasion condition of inspection track through set up a large amount of personnel of patrolling and examining nationwide, but comparatively consumes manpower and financial resources to, not rapid enough to urgent accident response.
In addition, video monitor system is used for railway safety monitoring, monitors whether there is the boundary of foreign matter invasion railway, however, present video monitor system needs the special messenger to monitor, and monitoring personnel's work load is very big, produces careless omission easily when tired, causes dangerous hidden danger. The existing method for monitoring the foreign matters by adopting the laser curtain wall has the advantages that the foreign matters are detected by installing a plurality of two-dimensional laser sensors to form the laser curtain wall, the method is high in detection speed and sensitivity, the installation is complex, the influence of the environment is large, only a few cross sections can be detected, and the cost is high.
Disclosure of Invention
The method is high in detection accuracy, labor-saving, automatic in alarming, low in cost and small in environmental influence.
In order to achieve the above object, the present application provides a method for identifying a foreign object in a railway based on machine learning, the method comprising the steps of: acquiring a monitoring video of a track in front of a train in the running process of the train, and acquiring a plurality of frames of monitoring images in the monitoring video in real time; establishing a dangerous area foreign matter detection window in the monitoring image, and extracting a track dangerous area detection image of the monitoring image in the dangerous area foreign matter detection window; preprocessing a track dangerous area detection image; inputting the preprocessed rail dangerous area detection image into a pre-established foreign matter identification classification model, identifying whether foreign matters exist in the rail dangerous area detection image or not by the foreign matter identification classification model, sending an alarm signal if the foreign matters exist in the rail dangerous area detection image, and otherwise, continuously identifying other rail dangerous area detection images.
As above, the method for establishing the foreign object identification classification model in advance comprises the following steps:
acquiring training image sets of various foreign matters;
respectively inputting the training image set of each foreign matter into a convolutional neural network for training to obtain a foreign matter identification submodel corresponding to the foreign matter;
and fusing the plurality of foreign matter recognition submodels to form a foreign matter recognition classification model.
The above method, wherein, for each foreign object, the foreign object image on the railway is collected in different weather and at different time, and from multiple angles, the foreign object contour image in the foreign object image is extracted, and the plurality of foreign object contour images form the training image set of the foreign object.
As above, the method for obtaining the foreign object recognition submodel includes:
inputting a training image set of foreign matters into an input layer of a convolutional neural network, and outputting an image matrix of the foreign matters by the input layer;
extracting the characteristics of the image matrix of the foreign matters from the convolution layer to obtain a characteristic image matrix of the foreign matters;
the pooling layer is used for pooling the output characteristic image matrix of the convolutional layer to obtain a key characteristic identification image matrix of the foreign matter;
inputting the key characteristic identification image matrix output by the pooling layer into a full-connection layer, and classifying the input key characteristic identification image matrix into a corresponding foreign matter type area by the full-connection layer;
and after the full connection layer classifies the characteristic image matrix into corresponding foreign matter type areas, the output layer outputs a foreign matter identification submodel of each type of foreign matter.
As described above, the calculation formula for obtaining the characteristic image matrix of the foreign object in the convolutional layer is:
Cij＝f(w·xj:j+h-1+b)；
wherein CijA characteristic image matrix which represents the j-th foreign matter with the type i corresponding to the characteristic image matrix after the convolution operation; f (-) represents the convolutional layer convolution kernel; w represents a weight matrix of the convolutional layer filter; h represents the length of the convolutional layer filter; b represents a bias; x is the number ofj:j+h-1And an image matrix representing the image of the j-th foreign object of the type i to the image matrix of the j + h-1-th foreign object.
As described above, the formula for pooling the characteristic image matrix output by the convolutional layer is as follows:
Figure BDA0002510745960000031
q is an odd number.
Wherein,
Figure BDA0002510745960000041
representing the characteristic value of a central pixel point of a window covering convolution layer output image with q multiplied by q dimension, and u representing the transverse coordinate of a pixel point of a window covering convolution layer output characteristic image with q multiplied by q dimension; v represents the vertical coordinate of the pixel point of the window covering convolution layer output characteristic image with q multiplied by q dimension; c (u, v) represents a characteristic value of a pixel point with q multiplied by q dimension window covering convolutional layer output characteristic image coordinates of (u, v).
As above, wherein the method of extracting the rail hazard zone detection image includes:
identifying the shape and position of the track in the monitored image;
establishing a dangerous area foreign matter detection window on the outer periphery of the track according to the shape and the position of the track;
and dividing the image positioned outside the dangerous area foreign matter detection window, reserving the image positioned inside the dangerous area foreign matter detection window, and forming a track dangerous area detection image.
As above, wherein, the method for preprocessing the rail dangerous area detection image comprises:
carrying out gray processing on the detection image of the rail dangerous area;
carrying out smooth filtering processing on the rail dangerous area detection image subjected to graying processing;
and calculating the definition of the detected image of the track dangerous area after the smooth filtering processing, judging whether the definition exceeds a preset qualified threshold value, if not, performing the definition processing on the image, otherwise, finishing the preprocessing.
As above, the calculation formula of the pixel value of the image output after the filtering process is as follows:
Figure BDA0002510745960000042
wherein, p (x, y) represents the pixel value of the pixel with the pixel coordinate (x, y) in the output image after the filtering processing, f (x, y) represents the pixel value of the pixel with the pixel coordinate (x, y) in the track dangerous area detection image, (x, y) represents the coordinate of the pixel, and sigma represents the variance; denotes a convolution operation; e is 2.718, k represents the dimension of the window.
The application also provides a railway foreign matter identification system based on machine learning, includes:
the image acquisition module is used for acquiring a monitoring video of a track in front of a train in the running process of the train and acquiring a plurality of frames of monitoring images in the monitoring video in real time;
the image extraction module is used for establishing a dangerous area foreign matter detection window in the monitoring image and extracting a track dangerous area detection image of the monitoring image in the dangerous area foreign matter detection window;
the image preprocessing module is used for preprocessing the detection image of the rail dangerous area;
and the foreign matter identification module is used for inputting the preprocessed rail dangerous area detection image into a pre-established foreign matter identification classification model, and the foreign matter identification classification model is used for identifying whether foreign matters exist in the rail dangerous area detection image or not.
The beneficial effect that this application realized is as follows:
(1) this application is through foreign matter discernment classification model, whether there is the foreign matter in the automatic identification image to the kind of output foreign matter, if discern the foreign matter, then report to the police automatically, remind navigating mate the place ahead to have dangerous foreign matter, prevent that danger from taking place.
(2) According to the method and the device, the foreign matter detection window in the dangerous area is established in the monitoring image, so that the influence of houses, plains or mountains and the like in the monitoring image on the monitoring result is eliminated, and the accuracy of foreign matter detection in the dangerous area of the track is improved.
(3) The method and the device for detecting the image sharpness detect whether the image sharpness meets the requirements after filtering processing, if so, the image sharpness is qualified, otherwise, the image sharpness is required, and the detection accuracy is improved.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments described in the present application, and other drawings can be obtained by those skilled in the art according to the drawings.
Fig. 1 is a flowchart of a railway foreign matter identification method based on machine learning according to the present application.
Fig. 2 is a schematic diagram of an image for detecting a dangerous area of a track according to the present application.
Fig. 3 is a flowchart of a method for extracting a track danger area detection image according to the present application.
Fig. 4 is a flowchart of a method for preprocessing a detection image of a dangerous area of a rail according to the present application.
Fig. 5 is a flowchart of a method for pre-establishing a foreign object identification classification model according to the present application.
FIG. 6 is a schematic structural diagram of a railway foreign object recognition system based on machine learning according to the present application
Reference numerals: 1-a dangerous area foreign matter detection window; 2-a first track; 3-a second track; 4-a foreign body; 10-an image acquisition module; 20-an image extraction module; 30-an image pre-processing module; 40-a foreign object identification module; 41-foreign object recognition classification model; 100-railway foreign body identification system.
Detailed Description
The technical solutions in the embodiments of the present application are clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some, but not all, embodiments of the present application. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
Example one
As shown in fig. 1, the present application provides a method for identifying a foreign object in a railway based on machine learning, which comprises the following steps:
and step S1, acquiring the monitoring video of the track in front of the train in the running process of the train, and acquiring multi-frame monitoring images in the monitoring video in real time.
Specifically, a plurality of time nodes are selected to obtain a plurality of frames of monitoring images of the monitoring video, each time node corresponds to one frame of monitoring image, and each frame of monitoring image reflects the environment condition of the track when the corresponding time node is located.
When the train runs at night, the front lighting system is started, and the condition that the monitoring video of the track in front of the train is collected is clear is guaranteed.
Step S2, a dangerous area foreign matter detection window is established in the monitoring image, and a track dangerous area detection image of the monitoring image in the dangerous area foreign matter detection window is extracted.
The foreign matter detection window in the dangerous area is established in the monitoring image, so that the influence of houses, plains or mountains and the like in the monitoring image on the monitoring result is eliminated, and the accuracy of foreign matter detection in the dangerous area of the track is improved.
As shown in fig. 3, step S2 includes the following sub-steps:
step S210, identify the shape and position of the track in the monitored image.
Step S220, according to the shape and position of the track, a dangerous area foreign object detection window is established on the outer periphery of the track, wherein the side of the dangerous area foreign object detection window is parallel to the track.
Step S230, dividing the image outside the dangerous region foreign matter detection window, and retaining the image inside the dangerous region foreign matter detection window to form a track dangerous region detection image.
Specifically, the foreign matter detection window in the danger area is established according to the shape of the rail, and if the railway rail is linear, the foreign matter detection window in the danger area of the linear railway rail is established; and if the railway track is in a curve shape, establishing a curve-shaped dangerous area foreign matter detection window. The side lines of the two sides of the dangerous area foreign matter detection window are parallel to the railway track line and offset to the two sides of the railway track line for a certain distance, the front edge of the dangerous area foreign matter detection window is positioned at the end of the railway track line, and the rear edge of the dangerous area foreign matter detection window is positioned at the edge close to the head of the train.
As shown in fig. 2, a hazardous area foreign matter detection window 1 is established, and both side edges of the hazardous area foreign matter detection window 1 are located at both sides of the first rail 2 and the second rail 3, so that a foreign matter 4 (as shown in fig. 2) can be detected in the hazardous area foreign matter detection window.
In step S3, the rail risk region detection image is preprocessed.
As shown in fig. 4, step S3 includes the following sub-steps:
in step S310, a grayscale process is performed on the rail risk region detection image.
The calculation formula for carrying out gray processing on the rail dangerous area detection image is as follows:
H＝0.3R+0.59G+0.11B；
h represents a gray value; r represents red; g represents green; b represents blue.
Step S320, performing smoothing filtering processing on the rail dangerous region detection image after the graying processing.
And carrying out smooth filtering treatment on the track dangerous area detection image, filtering high-frequency noise, restraining the noise of the track dangerous area detection image under the condition of keeping image detail characteristics as much as possible, and extracting the characteristics of the track dangerous area detection image, wherein the characteristics comprise contours or edges.
According to one embodiment of the invention, a k-dimension-multiplied k-dimension window containing a weighting coefficient is placed on the track dangerous area detection image, the track dangerous area detection image is weighted and averaged, the value of each pixel point is obtained by carrying out weighted average on the pixel point and other pixel values in the neighborhood of the window, namely, each pixel point in the window scanning image replaces the value of the central pixel point of the template by the weighted average value of the pixels in the neighborhood determined by the window.
According to an embodiment of the present invention, the calculation formula of the pixel values of the image output after the filtering process is as follows:
Figure BDA0002510745960000091
p (x, y) represents a pixel value of a pixel with a pixel coordinate (x, y) in an output image after filtering, f (x, y) represents a pixel value of a pixel with a pixel coordinate (x, y) in a track dangerous region detection image, (x, y) represents a coordinate of a pixel, sigma represents variance, and the larger sigma is, the wider frequency band is, and noise can be well suppressed; denotes a convolution operation; e denotes a constant, e is 2.718, and k denotes the dimension of the window.
Step S330, calculating the definition Q of the detection image of the track dangerous area after the smooth filtering processing, and judging whether the definition exceeds a preset qualified threshold QQualifiedAnd if not, performing sharpening processing on the image, otherwise, finishing the preprocessing.
Because the image after the image filtering processing can become fuzzy, whether the definition of the image after the filtering processing meets the requirement is detected, if so, the image is qualified, otherwise, the image needs to be subjected to the definition processing, and the detection accuracy is improved.
The calculation formula of the definition of the detected image of the track dangerous area after the smoothing filtering processing is as follows:
Figure BDA0002510745960000092
wherein Q represents definition, and p (x, y) represents a pixel value of a pixel with pixel point coordinates (x, y) in an output image after filtering processing; n represents the total row number of the output image matrix after filtering processing; m represents the total number of columns of the output image matrix after the filtering process.
Step S340, the track dangerous area detection image is subjected to sharpening processing, and an ideal and clear track dangerous area detection image is obtained.
The formula of the sharpening process is as follows:
Figure BDA0002510745960000101
wherein, T (x, y) represents the pixel value of the pixel with the pixel point coordinate (x, y) in the image after the sharpening processing, and J (r) represents the diffraction limit value of the actual lens; w represents frequency; f-1(. cndot.) denotes the inverse fourier operation.
And step S4, inputting the preprocessed rail dangerous area detection image into a pre-established foreign matter identification classification model, identifying whether foreign matters exist in the rail dangerous area detection image or not by the foreign matter identification classification model, sending an alarm signal if the foreign matters exist in the rail dangerous area detection image, and otherwise, continuously identifying other rail dangerous area detection images.
Specifically, after a preprocessed rail dangerous region detection image is input into a foreign matter identification classification model, the foreign matter identification classification model extracts a characteristic image matrix in the rail dangerous region detection image, compares the extracted characteristic image matrix with a key characteristic identification image matrix of a foreign matter stored in the rail dangerous region detection image, and judges that the foreign matter exists in the rail dangerous region detection image if the similarity between the extracted characteristic image matrix and the key characteristic identification image matrix of the foreign matter is greater than a preset threshold value, and obtains the type of the foreign matter with the maximum similarity value between the extracted characteristic image matrix and the key characteristic identification image matrix of a certain type of foreign matter as the type of the foreign matter existing in the rail dangerous region detection image; and if the similarity of the extracted characteristic image matrix and the key characteristic identification image matrix of the foreign matters is less than or equal to a preset threshold value, judging that the foreign matters do not exist in the detection image of the track dangerous area.
The foreign matter identification and classification model comprises key characteristic identification images of different types of foreign matters and is used for identifying the foreign matters and the types of the foreign matters.
As shown in fig. 5, the method for establishing the foreign object identification classification model in advance includes:
step S410, acquiring training image sets of various foreign matters.
Specifically, a foreign matter image causing an accident of the railway train is selected for training, the foreign matter image on the railway is collected from multiple angles at different weather and different time aiming at each foreign matter, the foreign matter outline image in the foreign matter image is extracted, and the plurality of foreign matter outline images form a training image set of the foreign matter.
The training image set of the foreign object is denoted X ═ Xi1，Xi2，…,Xin}; wherein i represents the type of foreign matter; n represents the total number of foreign body images acquired by each foreign body; xinAnd the n foreign body outline image represents the acquired i foreign body.
Step S420, inputting the training image set of each foreign matter into a convolutional neural network for training, and obtaining a foreign matter identification submodel corresponding to the foreign matter.
The convolutional neural network comprises an input layer, a convolutional layer, a pooling layer, a full-link layer and an output layer. Wherein, the convolution layer is a foreign matter feature extraction layer, and the feature of the foreign matter is extracted through a filter; the pooling layer is a feature mapping layer, and the features obtained after the convolution layer are sampled to obtain a local optimal value.
Specifically, step S420 includes the following sub-steps:
in step S421, a training image set of a foreign object is input to the input layer of the convolutional neural network, and the input layer outputs an image matrix of the foreign object.
In step S422, the features of the image matrix of the foreign object are extracted from the convolution layer, and a feature image matrix of the foreign object is obtained.
After the input layer outputs the image matrix, inputting the image matrix output by the input layer into the convolutional layer, and extracting a characteristic image matrix of the foreign matter in the convolutional layer;
the calculation formula for acquiring the characteristic image matrix of the foreign matter on the convolutional layer is as follows:
Cij＝f(w·xj:j+n-1+b)；
wherein CijA characteristic image matrix which represents the j-th foreign matter with the type i corresponding to the characteristic image matrix after the convolution operation; f (-) represents the convolutional layer convolution kernel; w represents a weight matrix of the convolutional layer filter; h represents the length of the convolutional layer filter; b represents a bias; x is the number ofj:j+h-1And an image matrix representing the image of the j-th foreign object of the type i to the image matrix of the j + h-1-th foreign object.
In step S423, the pooling layer pools the feature image matrix outputted from the convolutional layer to obtain a key feature identification image matrix of the foreign object.
Specifically, a q-by-q dimensional (i.e., q rows and q columns) window is established in the pooling layer, the q-by-q dimensional window is placed on the characteristic image output by the convolutional layer, the characteristic values of all pixel points in the window are summed, and then the average value is calculated to serve as the characteristic value of the central pixel point of the window, so that a key characteristic identification image matrix is obtained in the pooling layer.
The formula for pooling the characteristic image matrix output by the convolutional layer is as follows:
Figure BDA0002510745960000121
q is an odd number.
Wherein,
Figure BDA0002510745960000122
a window representing q times q dimensions covers the characteristic values of the central pixel points of the convolutional layer output image,
Figure BDA0002510745960000123
representing the coordinates of the center pixel of the q-by-q-dimensional window-covering convolution layer output image, u represents the q-by-q-dimensional window-covering convolutionOutputting the horizontal coordinates of the pixel points of the characteristic image in a lamination mode; v represents the vertical coordinate of the pixel point of the window covering convolution layer output characteristic image with q multiplied by q dimension; c (u, v) represents a characteristic value of a pixel point with q multiplied by q dimension window covering convolutional layer output characteristic image coordinates of (u, v).
Step S424, the key feature recognition image matrix output by the pooling layer is input to the full connection layer, and the full connection layer classifies the input key feature recognition image matrix into the corresponding foreign matter type region.
In step S425, the full link layer classifies the feature image matrix into corresponding foreign object type regions, and then the output layer outputs a foreign object identification submodel for each type of foreign object.
The foreign matter identifier model of each foreign matter type comprises a key feature identification image matrix of the foreign matter type.
And step S430, fusing or combining the plurality of foreign matter identification submodels together to form a foreign matter identification classification model.
Example two
As shown in fig. 5, the present application provides a machine learning based railway foreign object recognition system 100, comprising:
the image acquisition module 10 is used for acquiring a monitoring video of a track in front of a train in the running process of the train and acquiring multi-frame monitoring images in the monitoring video in real time;
the image extraction module 20 is configured to establish a dangerous region foreign matter detection window in the monitoring image, and extract a track dangerous region detection image of the monitoring image located in the dangerous region foreign matter detection window;
the image preprocessing module 30 is used for preprocessing the detection image of the rail dangerous area;
and the foreign matter identification module 40 is used for inputting the preprocessed track dangerous area detection image into a pre-established foreign matter identification classification model 41, wherein the foreign matter identification classification model 41 is used for identifying whether foreign matters exist in the track dangerous area detection image or not, if the foreign matters exist, an alarm signal is sent out, and otherwise, other track dangerous area detection images are continuously identified.
The foreign object recognition module 40 includes a foreign object recognition and classification model 41, and the foreign object recognition and classification model 41 is used for recognizing foreign objects and the types of the foreign objects.
The beneficial effect that this application realized is as follows:
(1) this application is through foreign matter discernment classification model, whether there is the foreign matter in the automatic identification image to the kind of output foreign matter, if discern the foreign matter, then report to the police automatically, remind navigating mate the place ahead to have dangerous foreign matter, prevent that danger from taking place.
(2) According to the method and the device, the foreign matter detection window in the dangerous area is established in the monitoring image, so that the influence of houses, plains or mountains and the like in the monitoring image on the monitoring result is eliminated, and the accuracy of foreign matter detection in the dangerous area of the track is improved.
(3) The method and the device for detecting the image sharpness detect whether the image sharpness meets the requirements after filtering processing, if so, the image sharpness is qualified, otherwise, the image sharpness is required, and the detection accuracy is improved.
The above description is only an embodiment of the present invention, and is not intended to limit the present invention. Various modifications and alterations to this invention will become apparent to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.

Claims (5)

1. A railway foreign matter identification method based on machine learning is characterized by comprising the following steps:
acquiring a monitoring video of a track in front of a train in the running process of the train, and acquiring a plurality of frames of monitoring images in the monitoring video in real time;
establishing a dangerous area foreign matter detection window in the monitoring image, and extracting a track dangerous area detection image of the monitoring image in the dangerous area foreign matter detection window;
preprocessing a track dangerous area detection image;
the method for preprocessing the rail dangerous area detection image comprises the following steps:
carrying out gray processing on the detection image of the rail dangerous area;
carrying out smooth filtering processing on the rail dangerous area detection image subjected to graying processing;
using a k-multiplied k-dimensional window containing a weighting coefficient to be placed on the track dangerous area detection image, carrying out weighted average on the track dangerous area detection image, wherein the value of each pixel point is obtained by carrying out weighted average on the value of each pixel point and other pixel values in the neighborhood of the window, namely scanning each pixel point in the image by using the window, and replacing the value of the central pixel point of the template by using the weighted average value of the pixels in the neighborhood determined by the window;
the calculation formula of the pixel value of the image output after the filtering process is as follows:
Figure FDA0003196659300000011
wherein, p (x, y) represents the pixel value of the pixel with the pixel coordinate (x, y) in the output image after the filtering processing, f (x, y) represents the pixel value of the pixel with the pixel coordinate (x, y) in the track dangerous area detection image, (x, y) represents the coordinate of the pixel, and sigma represents the variance; denotes a convolution operation; e is 2.718, k represents the dimension of the window;
calculating the definition of the detected image in the rail dangerous area after the smooth filtering processing, judging whether the definition exceeds a preset qualified threshold value, if not, performing the definition processing on the image, otherwise, finishing the preprocessing;
the calculation formula of the definition of the track dangerous area detection image after the smooth filtering processing is as follows:
Figure FDA0003196659300000021
wherein Q represents definition, and p (x, y) represents a pixel value of a pixel with pixel point coordinates (x, y) in an output image after filtering processing; n represents the total row number of the output image matrix after filtering processing; m represents the total column number of the output image matrix after filtering processing;
inputting the preprocessed rail dangerous area detection image into a pre-established foreign matter identification classification model, identifying whether foreign matters exist in the rail dangerous area detection image or not by the foreign matter identification classification model, if the foreign matters exist, sending an alarm signal, and if not, continuously identifying other rail dangerous area detection images;
in the process of pre-establishing a foreign matter identification classification model, selecting a foreign matter image causing an accident of a railway train for training, acquiring the foreign matter image on the railway from multiple angles at different days and different moments aiming at each foreign matter, extracting a foreign matter outline image in the foreign matter image, and forming a training image set of the foreign matter by the foreign matter outline images; respectively inputting the training image set of each foreign matter into a convolutional neural network for training to obtain a foreign matter identification submodel corresponding to the foreign matter;
wherein, the training image set of each foreign matter is respectively input into the convolutional neural network for training, which comprises the following steps: respectively and sequentially inputting the training image set of each foreign matter into an input layer, a convolution layer, a pooling layer, a full-link layer and an output layer of a convolutional neural network to obtain a foreign matter identification submodel;
the calculation formula for acquiring the characteristic image matrix of the foreign matter in the convolutional layer is as follows:
Cij＝f(w·xj：j+h-1+b)；
wherein CijA characteristic image matrix which represents the j-th foreign matter with the type i corresponding to the characteristic image matrix after the convolution operation; f (-) represents the convolutional layer convolution kernel; w represents a weight matrix of the convolutional layer filter; h represents the length of the convolutional layer filter; b represents a bias; x is the number ofj：j+h-1An image matrix representing an image of the j-th foreign object of the type i to the j + h-1-th foreign object;
the formula for pooling the characteristic image matrix output by the convolutional layer is as follows:
Figure FDA0003196659300000031
q is an odd number;
wherein,
Figure FDA0003196659300000032
representing the characteristic value of a central pixel point of a window covering convolution layer output image with q multiplied by q dimension, and u representing the transverse coordinate of a pixel point of a window covering convolution layer output characteristic image with q multiplied by q dimension; v represents the vertical coordinate of the pixel point of the window covering convolution layer output characteristic image with q multiplied by q dimension; c (u, v) represents a characteristic value of a pixel point with q multiplied by q dimension window covering convolutional layer output characteristic image coordinates of (u, v).
2. The machine learning-based railway foreign matter identification method according to claim 1, wherein the method for establishing the foreign matter identification classification model in advance comprises the following steps:
acquiring training image sets of various foreign matters;
respectively inputting the training image set of each foreign matter into a convolutional neural network for training to obtain a foreign matter identification submodel corresponding to the foreign matter;
and fusing the plurality of foreign matter recognition submodels to form a foreign matter recognition classification model.
3. The machine learning-based railway foreign matter identification method according to claim 2, wherein the method for obtaining the foreign matter identification submodel is as follows:
inputting a training image set of foreign matters into an input layer of a convolutional neural network, and outputting an image matrix of the foreign matters by the input layer;
extracting the characteristics of the image matrix of the foreign matters from the convolution layer to obtain a characteristic image matrix of the foreign matters;
the pooling layer is used for pooling the output characteristic image matrix of the convolutional layer to obtain a key characteristic identification image matrix of the foreign matter;
inputting the key characteristic identification image matrix output by the pooling layer into a full-connection layer, and classifying the input key characteristic identification image matrix into a corresponding foreign matter type area by the full-connection layer;
and after the full connection layer classifies the characteristic image matrix into corresponding foreign matter type areas, the output layer outputs a foreign matter identification submodel of each type of foreign matter.
4. The machine learning-based railway foreign matter identification method according to claim 1, wherein the method of extracting a rail risk region detection image comprises:
identifying the shape and position of the track in the monitored image;
establishing a dangerous area foreign matter detection window on the outer periphery of the track according to the shape and the position of the track;
and dividing the image positioned outside the dangerous area foreign matter detection window, reserving the image positioned inside the dangerous area foreign matter detection window, and forming a track dangerous area detection image.
5. A railway foreign object recognition system based on machine learning, comprising:
the image acquisition module is used for acquiring a monitoring video of a track in front of a train in the running process of the train and acquiring a plurality of frames of monitoring images in the monitoring video in real time;
the image extraction module is used for establishing a dangerous area foreign matter detection window in the monitoring image and extracting a track dangerous area detection image of the monitoring image in the dangerous area foreign matter detection window;
the image preprocessing module is used for preprocessing the detection image of the rail dangerous area;
the method for preprocessing the rail dangerous area detection image comprises the following steps:
carrying out gray processing on the detection image of the rail dangerous area;
carrying out smooth filtering processing on the rail dangerous area detection image subjected to graying processing;
wherein the smoothing filter process includes: placing a k-multiplied k-dimensional window containing a weighting coefficient on the track dangerous area detection image, carrying out weighted average on the track dangerous area detection image, wherein the value of each pixel point is obtained by carrying out weighted average on the value of each pixel point and other pixel values in the neighborhood of the window, namely scanning each pixel point in the image by using the window, and replacing the value of the central pixel point of the template by using the weighted average value of the pixels in the neighborhood determined by the window;
the calculation formula of the pixel value of the image output after the filtering process is as follows:
Figure FDA0003196659300000051
wherein, p (x, y) represents the pixel value of the pixel with the pixel coordinate (x, y) in the output image after the filtering processing, f (x, y) represents the pixel value of the pixel with the pixel coordinate (x, y) in the track dangerous area detection image, (x, y) represents the coordinate of the pixel, and sigma represents the variance; denotes a convolution operation; e is 2.718, k represents the dimension of the window;
calculating the definition of the detected image in the rail dangerous area after the smooth filtering processing, judging whether the definition exceeds a preset qualified threshold value, if not, performing the definition processing on the image, otherwise, finishing the preprocessing;
the calculation formula of the definition of the track dangerous area detection image after the smooth filtering processing is as follows:
Figure FDA0003196659300000052
wherein Q represents definition, and p (x, y) represents a pixel value of a pixel with pixel point coordinates (x, y) in an output image after filtering processing; n represents the total row number of the output image matrix after filtering processing; m represents the total column number of the output image matrix after filtering processing;
the foreign matter identification module is used for inputting the preprocessed rail dangerous area detection image into a pre-established foreign matter identification classification model, and the foreign matter identification classification model identifies whether foreign matters exist in the rail dangerous area detection image or not;
in the process of pre-establishing a foreign matter identification classification model, selecting a foreign matter image causing an accident of a railway train for training, acquiring the foreign matter image on the railway from multiple angles at different days and different moments aiming at each foreign matter, extracting a foreign matter outline image in the foreign matter image, and forming a training image set of the foreign matter by the foreign matter outline images; respectively inputting the training image set of each foreign matter into a convolutional neural network for training to obtain a foreign matter identification submodel corresponding to the foreign matter;
wherein, the training image set of each foreign matter is respectively input into the convolutional neural network for training, which comprises the following steps: respectively and sequentially inputting the training image set of each foreign matter into an input layer, a convolution layer, a pooling layer, a full-link layer and an output layer of a convolutional neural network to obtain a foreign matter identification submodel;
the calculation formula for acquiring the characteristic image matrix of the foreign matter in the convolutional layer is as follows:
Cij＝f(w·xj：j+h-1+b)；
wherein CijA characteristic image matrix which represents the j-th foreign matter with the type i corresponding to the characteristic image matrix after the convolution operation; f (-) represents the convolutional layer convolution kernel; w represents a weight matrix of the convolutional layer filter; h represents the length of the convolutional layer filter; b represents a bias; x is the number ofj：j+h-1An image matrix representing an image of the j-th foreign object of the type i to the j + h-1-th foreign object;
the formula for pooling the characteristic image matrix output by the convolutional layer is as follows:
Figure FDA0003196659300000061
q is an odd number;
wherein,
Figure FDA0003196659300000071
representing the characteristic value of a central pixel point of a window covering convolution layer output image with q multiplied by q dimension, and u representing the transverse coordinate of a pixel point of a window covering convolution layer output characteristic image with q multiplied by q dimension; v represents the vertical coordinate of the pixel point of the window covering convolution layer output characteristic image with q multiplied by q dimension; c (u, v) represents q times q dimension window covering convolution layer output characteristic image coordinate as (u, v) of the pixel points.
CN202010460331.9A 2020-05-27 2020-05-27 Railway foreign matter identification method and system based on machine learning Active CN111626203B (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN202010460331.9A CN111626203B (en) 2020-05-27 2020-05-27 Railway foreign matter identification method and system based on machine learning
ZA2020/04403A ZA202004403B (en) 2020-05-27 2020-07-17 A railway foreign object recognition method and system based on machine learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010460331.9A CN111626203B (en) 2020-05-27 2020-05-27 Railway foreign matter identification method and system based on machine learning

Publications (2)

Publication Number Publication Date
CN111626203A CN111626203A (en) 2020-09-04
CN111626203B true CN111626203B (en) 2021-11-16

Family

ID=72272139

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010460331.9A Active CN111626203B (en) 2020-05-27 2020-05-27 Railway foreign matter identification method and system based on machine learning

Country Status (2)

Country Link
CN (1) CN111626203B (en)
ZA (1) ZA202004403B (en)

Families Citing this family (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112508893B (en) * 2020-11-27 2024-04-26 中国铁路南宁局集团有限公司 Method and system for detecting tiny foreign matters between double rails of railway based on machine vision
CN112722009A (en) * 2021-01-20 2021-04-30 李林卿 Danger avoiding method and device for railway transportation and train-mounted terminal
CN113553916B (en) * 2021-06-30 2023-04-07 广西大学 A Convolutional Neural Network Based Obstacle Detection Method for Track Dangerous Areas
CN113779734A (en) * 2021-09-22 2021-12-10 重庆华渝重工机电有限公司 Straddle type single-track turnout monitoring and maintaining system based on artificial intelligence
CN114565886A (en) * 2021-12-16 2022-05-31 高新兴创联科技有限公司 Method for detecting dangerous behavior of personnel crossing station tracks in railway industry
CN114355467A (en) * 2021-12-30 2022-04-15 宁波微科光电股份有限公司 A method for establishing a detection model and a multi-image-based orbital foreign object detection method
CN114898204B (en) * 2022-03-03 2023-09-05 中国铁路设计集团有限公司 Rail transit peripheral dangerous source detection method based on deep learning
CN114782828B (en) * 2022-06-22 2022-09-09 国网山东省电力公司高青县供电公司 Foreign matter detection system based on deep learning
CN115205796B (en) * 2022-07-07 2023-05-16 北京交通大学 Rail line foreign matter intrusion monitoring and risk early warning method and system
CN117994753B (en) * 2024-04-03 2024-06-07 浙江浙能数字科技有限公司 Vision-based device and method for detecting abnormality of entrance track of car dumper

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108197610A (en) * 2018-02-02 2018-06-22 北京华纵科技有限公司 A kind of track foreign matter detection system based on deep learning
CN109188460A (en) * 2018-09-25 2019-01-11 北京华开领航科技有限责任公司 Unmanned foreign matter detection system and method
CN109446913A (en) * 2018-09-28 2019-03-08 桂林电子科技大学 A kind of detection method for judging vehicle bottom and whether reequiping
CN109461163A (en) * 2018-07-20 2019-03-12 河南师范大学 A kind of edge detection extraction algorithm for magnetic resonance standard water mould
CN109649091A (en) * 2018-12-28 2019-04-19 泉州装备制造研究所 Monitoring system for tyres of automobile based on computer vision
CN110532889A (en) * 2019-08-02 2019-12-03 南京理工大学 Track foreign matter detecting method based on rotor unmanned aircraft and YOLOv3
CN111191559A (en) * 2019-12-25 2020-05-22 国网浙江省电力有限公司泰顺县供电公司 Overhead line early warning system obstacle identification method based on time convolution neural network

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101488219B (en) * 2008-12-19 2010-12-22 四川虹微技术有限公司 Fast video image bilateral filtering method
CN103927749A (en) * 2014-04-14 2014-07-16 深圳市华星光电技术有限公司 Image processing method and device and automatic optical detector
CN104850836B (en) * 2015-05-15 2018-04-10 浙江大学 Insect automatic distinguishing method for image based on depth convolutional neural networks
CN108416745B (en) * 2018-02-02 2020-06-26 中国科学院西安光学精密机械研究所 Image self-adaptive defogging enhancement method with color constancy
CN110428475B (en) * 2019-06-21 2021-02-05 腾讯科技（深圳）有限公司 Medical image classification method, model training method and server

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108197610A (en) * 2018-02-02 2018-06-22 北京华纵科技有限公司 A kind of track foreign matter detection system based on deep learning
CN109461163A (en) * 2018-07-20 2019-03-12 河南师范大学 A kind of edge detection extraction algorithm for magnetic resonance standard water mould
CN109188460A (en) * 2018-09-25 2019-01-11 北京华开领航科技有限责任公司 Unmanned foreign matter detection system and method
CN109446913A (en) * 2018-09-28 2019-03-08 桂林电子科技大学 A kind of detection method for judging vehicle bottom and whether reequiping
CN109649091A (en) * 2018-12-28 2019-04-19 泉州装备制造研究所 Monitoring system for tyres of automobile based on computer vision
CN110532889A (en) * 2019-08-02 2019-12-03 南京理工大学 Track foreign matter detecting method based on rotor unmanned aircraft and YOLOv3
CN111191559A (en) * 2019-12-25 2020-05-22 国网浙江省电力有限公司泰顺县供电公司 Overhead line early warning system obstacle identification method based on time convolution neural network

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
铁路轨道异物入侵的智能识别研究;李丹丹;《中国优秀硕士学位论文全文数据库 工程科技II辑》;20170415;正文第6-43页 *

Also Published As

Publication number Publication date
ZA202004403B (en) 2021-03-31
CN111626203A (en) 2020-09-04

Similar Documents

Publication Publication Date Title
CN111626203B (en) Railway foreign matter identification method and system based on machine learning
CN111626204B (en) Railway foreign matter invasion monitoring method and system
CN111626207B (en) Method and system for detecting intrusion of foreign matters in front of train based on image processing
KR101748121B1 (en) System and method for detecting image in real-time based on object recognition
CN110261436A (en) Rail deformation detection method and system based on infrared thermal imaging and computer vision
CN111275910B (en) Method and system for detecting border crossing behavior of escalator based on Gaussian mixture model
EP0567059A1 (en) Object recognition system and abnormality detection system using image processing
CN103077526B (en) There is train method for detecting abnormality and the system of depth detection function
US20100177963A1 (en) Situation determining apparatus, situation determining method, situation determining program, abnormality determining apparatus, abnormality determining method, abnormality determining program, and congestion estimating apparatus
CN110862033A (en) Intelligent early warning detection method applied to coal mine inclined shaft winch
CN111626170B (en) Image recognition method for railway side slope falling stone intrusion detection
CN112327698A (en) Flood disaster early warning system and method based on Internet of things
CN111783700B (en) Automatic recognition and early warning method and system for pavement foreign matters
CN111259718A (en) Escalator retention detection method and system based on Gaussian mixture model
CN113484858A (en) Intrusion detection method and system
CN101739549A (en) Face detection method and system
WO2022267266A1 (en) Vehicle control method based on visual recognition, and device
CN104182983A (en) Highway monitoring video definition detection method based on corner features
CN111523386A (en) Machine vision-based high-speed railway platform door monitoring and protecting method and system
CN113076821A (en) Event detection method and device
CN113537007B (en) Non-staff intrusion detection and alarm method and device applied to railway platform
CN115205796A (en) Method and system for foreign body intrusion limit monitoring and risk early warning in track lines
CN113553916A (en) A Convolutional Neural Network-based Obstacle Detection Method in Track Hazardous Areas
CN108109146A (en) A kind of pavement marker line defect detection device
CN116597394A (en) Railway foreign matter intrusion detection system and method based on deep learning

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
GR01 Patent grant
GR01 Patent grant
TR01 Transfer of patent right
TR01 Transfer of patent right

Effective date of registration: 20230711

Address after: Room 1503, Hongqilou Business Building, No. 11, Gongye East Street, Qiaodong District, Zhangjiakou City, Hebei Province, 075000

Patentee after: Zhangjiakou Dongchu Technology Co.,Ltd.

Address before: 101300 room 3001, 3rd floor, 102 door, building 8, yard 12, Xinzhong street, Nanfaxin Town, Shunyi District, Beijing

Patentee before: Beijing Weijie Dongbo Information Technology Co.,Ltd.

EE01 Entry into force of recordation of patent licensing contract
EE01 Entry into force of recordation of patent licensing contract

Application publication date: 20200904

Assignee: Hebei Yuncha Technology Co.,Ltd.

Assignor: Zhangjiakou Dongchu Technology Co.,Ltd.

Contract record no.: X2023980049711

Denomination of invention: A Machine Learning Based Railway Foreign Object Recognition Method and System

Granted publication date: 20211116

License type: Common License

Record date: 20231204