Methods and Systems for Identifying Abnormal Passenger Behavior in Urban Rail Stations

Abstract

This invention provides a method and system for identifying abnormal passenger behavior in urban rail transit stations, belonging to the field of urban rail transit operation and maintenance technology. The method involves acquiring surveillance video of the urban rail transit station area; using a pre-trained detection model to detect passengers in the acquired video, obtaining key point information of passengers in the video; calculating the key human motion features of each passenger based on the acquired key point information using a motion key feature calculation method; and classifying the key human motion features of passengers using a pre-trained random forest model to obtain the current behavior identification result of the passengers. This invention utilizes surveillance video to identify abnormal passenger behavior in urban rail transit station areas, improving identification accuracy and reducing the workload of station operators through information technology, which is of great significance for ensuring the operational safety of urban rail transit stations.

Description

Technical Field

[0001] This invention relates to the field of urban rail transit operation and maintenance technology, specifically to a method and system for identifying abnormal passenger behavior in urban rail stations based on OpenPose. Background Art

[0002] Safety is the prerequisite and foundation for urban rail transit to fulfill its fundamental tasks and missions, especially for passengers. Stations are a major activity space for passengers within the urban rail transit system. Urban rail stations are relatively enclosed spaces, and most are built underground. Once passengers exhibit abnormal behavior or actions, it may trigger major safety incidents such as the butterfly effect. Therefore, it is essential to identify abnormal passenger behavior in urban rail stations.

[0003] Existing methods employ continuous adaptive mean shift, introduce corrected background weight histograms and odorless particle filtering techniques to handle interference from occlusion and similar color objects. Zhu Xudong applied Hidden Markov Models to abnormal behavior detection, proposing a novel framework for automatic behavior modeling and online anomaly detection that does not require manually labeled training datasets. Shah et al. used an improved human detection algorithm for human detection, then extracted geometric features of the human skeleton using a feature extraction algorithm, and completed the classification task using binary support vector machines and multi-class support vector machines. Chen Yufeng et al. proposed an abnormal behavior recognition method that combines a Bayesian framework with support vector machine methods, using a hybrid approach to track the human body and then classifying action types using support vector machines based on body structure and historical motion image information. Sun Xian proposed a passenger abnormal behavior recognition method that first constructs a temporal matrix, then constructs behavioral unit pairs for action classification, and finally uses sparse encoded histograms and sparse reconstruction to locate abnormal behavior after obtaining the classification results. Lu Ying designed a method to classify and identify four unsafe behaviors of subway passengers—smoking, punching, and waving for help with one or both arms—using Kinect skeletal point information to construct a motion feature vector with the maximum angle difference as the feature vector, and using a dynamic time warping algorithm.

[0004] However, the methods used to solve various tasks cannot be generalized. Different considerations are needed for specific tasks and application scenarios. For the urban rail station scenario, the existing research mentioned above uses surveillance video as the information acquisition device. There is no mature research and methodology system for this scenario, and the identification of abnormal passenger behavior in the acquired video is still imperfect. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for identifying abnormal passenger behavior in urban rail stations, so as to solve at least one of the technical problems existing in the background art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] On one hand, the present invention provides a method for identifying abnormal passenger behavior in urban rail transit stations, which includes:

[0008] Obtain surveillance video of the urban rail station area;

[0009] Using a pre-trained detection model, passenger detection is performed on the acquired surveillance video of the urban rail station area to obtain key point information of passengers in the video;

[0010] Based on the acquired passenger key point information, the key human motion features of each passenger are calculated according to the motion key feature calculation method.

[0011] Using a pre-trained random forest model, the key features of passengers' human body movements are classified to obtain the current behavior identification results of passengers.

[0012] Preferably, the key information of passengers in the video includes:

[0013] Features of the input image are extracted using a VGG-19 network, and the extracted feature maps are then fed into two convolutional neural networks.

[0014] The confidence level and local affinity field of human key points are calculated in the two convolutional neural networks respectively to realize the localization and preliminary association of key points;

[0015] Based on the confidence level and local affinity field of human key points, the Hungarian algorithm is used to solve the final key point connection relationship by utilizing the weight of the validity of the key point connection relationship.

[0016] Preferably, the specific formula for calculating the confidence level is as follows:

[0017]

[0018]

[0019] Where, x i,j Let σ be the true coordinates of keypoint i of individual j in the image, σ be an empirical value, and p be the coordinates of a point in the image.

[0020] The specific formula for calculating the local affinity field is as follows:

[0021]

[0022] Where, n c (p) represents the number of overlapping limbs c at point p;

[0023]

[0024] in, These are the coordinates of two key points that make up the limb c of the j-th individual.

[0025] Preferably, the specific formula for calculating the weight of the validity of the key point connection relationship is as follows:

[0026]

[0027] in p is the key point A point on the line.

[0028] Preferably, the specific calculation formula for the key features of the step is as follows:

[0029]

[0030]

[0031]

[0032]

[0033]

[0034]

[0035]

[0036]

[0037]

[0038]

[0039]

[0040]

[0041]

[0042]

[0043]

[0044]

[0045] Where x i Let y be the x-coordinate of key point i. i Let L be the ordinate of key point i. i-j Let be the distance between keypoint i and keypoint j.

[0046] Preferably, the random forest model classification process includes the following steps:

[0047] For the training set, the bootstrap method is used to extract a new training set of size m with replacement for each decision tree; for m features in the training set, n features are randomly selected at each non-leaf node of each tree, and the feature with the strongest classification ability is selected from these n features by evaluation indicators such as information gain ratio; each decision tree is recursively constructed, and no pruning is performed on the constructed decision trees; all the generated decision trees are combined into a random forest, and the probability of the input data being each category is obtained by the random forest, and the final predicted category is determined by voting.

[0048] Secondly, the present invention provides a system for identifying abnormal passenger behavior at urban rail transit stations, comprising:

[0049] The acquisition module is used to acquire surveillance video of the urban rail station area;

[0050] The detection module is used to detect passengers in the acquired surveillance video of the urban rail station area using a pre-trained detection model, and obtain key point information of passengers in the video.

[0051] The calculation module is used to calculate the key human motion features of each passenger based on the acquired passenger key point information and according to the motion key feature calculation method.

[0052] The classification module uses a pre-trained random forest model to classify the key features of passengers' human body movements and obtain the current behavior identification results of passengers.

[0053] Thirdly, the present invention provides a non-transitory computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the method for identifying abnormal passenger behavior at urban rail stations as described above.

[0054] Fourthly, the present invention provides a computer program product, including a computer program that, when run on one or more processors, is used to implement the method for identifying abnormal passenger behavior at urban rail stations as described above.

[0055] Fifthly, the present invention provides an electronic device, comprising: a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions for implementing the method for identifying abnormal passenger behavior at urban rail stations as described above.

[0056] The beneficial effects of this invention are: it utilizes surveillance video to identify abnormal passenger behavior in urban rail station areas, and improves the accuracy of identification by using information technology to identify abnormal passenger behavior within the station, thereby reducing the workload of station operators and playing a significant role in ensuring the operational safety of urban rail stations.

[0057] The advantages of additional aspects of the invention will be set forth more clearly in the following description or will be learned by practice of the invention. Attached Figure Description

[0058] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0059] Figure 1 This is a flowchart illustrating the passenger abnormal behavior identification process according to an embodiment of the present invention.

[0060] Figure 2 This is a diagram showing the location of key points as described in an embodiment of the present invention.

[0061] Figure 3 This is a schematic diagram illustrating the key points of the throwing behavior described in the embodiments of the present invention.

[0062] Figure 4 This is a schematic diagram illustrating the key points of concern in the fighting behavior described in the embodiments of the present invention.

[0063] Figure 5 This is a schematic diagram illustrating the key points of concern regarding the falling behavior described in an embodiment of the present invention.

[0064] Figure 6 The diagram illustrates the effect of an embodiment of the present invention. Detailed Implementation

[0065] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0066] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0067] It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with their meanings in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless defined as here.

[0068] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, and / or groups thereof.

[0069] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0070] To facilitate understanding of the present invention, the present invention will be further explained and described below with reference to the accompanying drawings and specific embodiments. However, the specific embodiments do not constitute a limitation on the embodiments of the present invention.

[0071] Those skilled in the art should understand that the accompanying drawings are merely schematic diagrams of embodiments, and the components in the drawings are not necessarily essential for implementing the present invention.

[0072] Example 1

[0073] This embodiment 1 provides a passenger abnormal behavior identification system for urban rail stations, including:

[0074] The acquisition module is used to acquire surveillance video of the urban rail station area;

[0075] The detection module is used to detect passengers in the acquired surveillance video of the urban rail station area using a pre-trained detection model, and obtain key point information of passengers in the video.

[0076] The calculation module is used to calculate the key human motion features of each passenger based on the acquired passenger key point information and according to the motion key feature calculation method.

[0077] The classification module uses a pre-trained random forest model to classify the key features of passengers' human body movements and obtain the current behavior identification results of passengers.

[0078] In this embodiment 1, the above-described system is used to implement a method for identifying abnormal passenger behavior in urban rail transit stations, including:

[0079] First, the acquisition module is used to obtain surveillance video of the urban rail station area;

[0080] Then, using the detection module, based on the pre-trained detection model, passenger detection is performed on the acquired surveillance video of the urban rail station area to obtain key point information of passengers in the video;

[0081] Next, using the calculation module, based on the acquired passenger key point information, the key human motion features of each passenger are calculated according to the motion key feature calculation method.

[0082] Finally, using the classification module, based on a pre-trained random forest model, the key features of the passenger's human body movements are classified to obtain the passenger's current behavior identification results.

[0083] The key information obtained from the video of the passenger includes:

[0084] Features of the input image are extracted using a VGG-19 network, and the extracted feature maps are then fed into two convolutional neural networks.

[0085] The confidence level and local affinity field of human key points are calculated in the two convolutional neural networks respectively to realize the localization and preliminary association of key points;

[0086] Based on the confidence level and local affinity field of human key points, the Hungarian algorithm is used to solve the final key point connection relationship by utilizing the weight of the validity of the key point connection relationship.

[0087] The specific formula for calculating the confidence level is as follows:

[0088]

[0089]

[0090] Where, x i,j Let σ be the true coordinates of keypoint i of individual j in the image, σ be an empirical value, and p be the coordinates of a point in the image.

[0091] The specific formula for calculating the local affinity field is as follows:

[0092]

[0093] Where, nc (p) represents the number of overlapping limbs c at point p;

[0094]

[0095] in, These are the coordinates of two key points that make up limb c of the j-th individual. The specific formula for calculating the weight of the validity of the connection relationship between these key points is as follows:

[0096]

[0097] in p is the key point A point on the line connecting the points. The specific calculation formula for the key features of the step is as follows:

[0098]

[0099]

[0100]

[0101]

[0102]

[0103]

[0104]

[0105]

[0106]

[0107]

[0108]

[0109]

[0110]

[0111]

[0112]

[0113]

[0114] Where x i Let y be the x-coordinate of key point i. i Let L be the ordinate of key point i. i-jLet be the distance between keypoint i and keypoint j.

[0115] The classification process of the random forest model includes the following steps:

[0116] For the training set, the bootstrap method is used to extract a new training set of size m with replacement for each decision tree; for m features in the training set, n features are randomly selected at each non-leaf node of each tree, and the feature with the strongest classification ability is selected from these n features by evaluation indicators such as information gain ratio; each decision tree is recursively constructed, and no pruning is performed on the constructed decision trees; all the generated decision trees are combined into a random forest, and the probability of the input data being each category is obtained by the random forest, and the final predicted category is determined by voting.

[0117] Example 2

[0118] In this embodiment 2, a method is provided to identify abnormal passenger behavior in different areas of an urban rail station using surveillance video, including the following steps:

[0119] Step 1: Obtain surveillance videos of various areas of the urban rail station;

[0120] Step 2: Input the acquired urban rail station surveillance video into the OpenPose model to detect passengers and obtain key point information of passengers in the video;

[0121] Step 3: Calculate the acquired passenger key point information according to the motion key feature calculation method to obtain the human motion key features of each passenger.

[0122] Step 4: Input the key features of each passenger's human body movements into the trained random forest model for classification and identification of the passenger's current behavior.

[0123] Step 2, which involves obtaining key information about passengers in the video, includes the following steps:

[0124] Step 21: Extract features from the input image using a VGG-19 network, and then input the extracted feature maps into two convolutional neural networks.

[0125] Step 22: Calculate the confidence score and local affinity field of human keypoints in both networks to achieve keypoint localization and preliminary association, and output the confidence score S of various keypoints. t and local affinity field L t ;

[0126] Step 23: Use the Hungarian algorithm to solve for the final key point connection relationship by utilizing the weight E of the validity of the key point connection relationship.

[0127] The specific formula for calculating the confidence level in step 22 is as follows:

[0128]

[0129]

[0130] Where, x i,j Let σ be the true coordinates of keypoint i of individual j in the image, σ be an empirical value, and p be the coordinates of a point in the image.

[0131] The specific formula for calculating the local affinity field is as follows:

[0132]

[0133] Where, n c (p) represents the number of overlapping limbs c at point p.

[0134]

[0135] in, These are the coordinates of two key points that make up the limb c of the j-th individual.

[0136] The specific formula for calculating the weight E of the validity of the key point connection relationship in step 23 is as follows:

[0137]

[0138] in p is the key point A point on the line.

[0139] The specific calculation formula for the key features of the action in step 3 is as follows:

[0140]

[0141]

[0142]

[0143]

[0144]

[0145]

[0146]

[0147]

[0148]

[0149]

[0150]

[0151]

[0152]

[0153]

[0154]

[0155]

[0156] Where x i Let y be the x-coordinate of key point i. i Let L be the ordinate of key point i. i-j Let be the distance between keypoint i and keypoint j.

[0157] Step 4, the random forest model classification process, includes the following steps:

[0158] Step 41: For the training set N, use the bootstrap method to extract a new training set of size N with replacement for each decision tree.

[0159] Step 42: For the m features in the training set, randomly select n features at each non-leaf node of each tree, and select the feature with the strongest classification ability from these n features using evaluation metrics such as information gain ratio.

[0160] Step 43: Recursively construct each decision tree, without pruning the completed decision trees.

[0161] Step 44: Combine all the generated decision trees into a random forest, use the random forest to obtain the probability of the input data being each category, and use a voting method to determine the final predicted category of the output.

[0162] In this second embodiment, a method for identifying abnormal passenger behavior in urban rail station areas is implemented using surveillance video. This allows urban rail stations to identify abnormal passenger behavior in daily operations through information technology, reducing the workload of station operators and playing a significant role in ensuring the operational safety of urban rail stations.

[0163] Example 3

[0164] like Figure 1 As shown in the figure, this invention discloses a method for identifying abnormal passenger behavior in urban rail transit stations based on OpenPose, including the following steps:

[0165] Step 1: Obtain surveillance videos of various areas of the urban rail station;

[0166] Step 2: Input the acquired urban rail station surveillance video into the OpenPose model to detect passengers and obtain key point information of passengers in the video. The locations of the key points are as follows: Figure 2 As shown;

[0167] Step 3: Calculate the acquired passenger key point information using the motion key feature calculation method to obtain the human motion key features for each passenger. For each behavior, the key points of interest are as follows: Figure 3 , Figure 4 , Figure 5 As shown;

[0168] Step 4: Input the key features of each passenger's human body movements into the trained random forest model for classification and identification of the passenger's current behavior.

[0169] In this embodiment, obtaining the key point information of the passenger in the video in step 2 includes the following steps:

[0170] Step 21: Extract features from the input image using a VGG-19 network, and then input the extracted feature maps into two convolutional neural networks.

[0171] Step 22: Calculate the confidence score and local affinity field of human keypoints in both networks to achieve keypoint localization and preliminary association, and output the confidence score S of various keypoints. t and local affinity field L t ;

[0172] Step 23: Use the Hungarian algorithm to solve for the final key point connection relationship by utilizing the weight E of the validity of the key point connection relationship.

[0173] The specific formula for calculating the confidence level in step 22 is as follows:

[0174]

[0175]

[0176] Where, x i,j Let σ be the true coordinates of keypoint i of individual j in the image, σ be an empirical value, and p be the coordinates of a point in the image.

[0177] The specific formula for calculating the local affinity field is as follows:

[0178]

[0179] Where, n c (p) represents the number of overlapping limbs c at point p.

[0180]

[0181] in, These are the coordinates of two key points that make up the limb c of the j-th individual.

[0182] The specific formula for calculating the weight E of the validity of the key point connection relationship in step 23 is as follows:

[0183]

[0184] in p is the key point A point on the line connecting the points. The specific calculation formula for the key features of the action in step 3 is as follows:

[0185]

[0186]

[0187]

[0188]

[0189]

[0190]

[0191]

[0192]

[0193]

[0194]

[0195]

[0196]

[0197]

[0198]

[0199]

[0200]

[0201] Where x i Let y be the x-coordinate of key point i. i Let L be the ordinate of key point i. i-j Let be the distance between keypoint i and keypoint j.

[0202] Step 4, the random forest model classification process, includes the following steps:

[0203] Step 41: For the training set N, use the bootstrap method to extract a new training set of size N with replacement for each decision tree.

[0204] Step 42: For the m features in the training set, randomly select n features at each non-leaf node of each tree, and select the feature with the strongest classification ability from these n features using evaluation metrics such as information gain ratio.

[0205] Step 43: Recursively construct each decision tree, without pruning the completed decision trees.

[0206] Step 44: Combine all the generated decision trees into a random forest, use the random forest to obtain the probability of the input data being each category, and use a voting method to determine the final predicted category of the output.

[0207] In step 41, the dataset used to train the random forest model consists of RGB video data of kicking, punching, falling, and throwing from the NTU RGB+D dataset released by Nanyang Technological University in Singapore, and self-shot samples. After obtaining the video samples, they are output as image samples at 5-frame intervals. Samples with similar actions and inconsistent with the target actions are initially removed. Skeletal point data is extracted from the obtained preliminary samples using the Open Pose model. Samples with too much missing skeletal point information and excessive noise are further removed, ultimately forming a dataset of 2500 samples each of walking, waiting, fighting, throwing, and falling during a normal ride, constituting a dataset of 10,000 data points.

[0208] In this embodiment 3, due to the specific nature of the application scenario, it is impossible to conduct the experiment in a real urban rail station. Therefore, simulated videos are used for verification. Simulated videos of various abnormal behaviors are used for detection. The effect is shown in the figure below. Figure 6 As shown in the figure. Experiments show that the accuracy of the present invention in identifying abnormal passenger behavior is higher than that of existing related technologies.

[0209] Example 4

[0210] Embodiment 4 of the present invention provides a non-transitory computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, a method for identifying abnormal passenger behavior in urban rail transit stations is implemented. The method includes:

[0211] Obtain surveillance video of the urban rail station area;

[0212] Using a pre-trained detection model, passenger detection is performed on the acquired surveillance video of the urban rail station area to obtain key point information of passengers in the video;

[0213] Based on the acquired passenger key point information, the key human motion features of each passenger are calculated according to the motion key feature calculation method.

[0214] Using a pre-trained random forest model, the key features of passengers' human body movements are classified to obtain the current behavior identification results of passengers.

[0215] Example 5

[0216] Embodiment 5 of the present invention provides a computer program (product), including a computer program that, when run on one or more processors, is used to implement a method for identifying abnormal passenger behavior in urban rail transit stations. The method includes:

[0217] Obtain surveillance video of the urban rail station area;

[0218] Using a pre-trained detection model, passenger detection is performed on the acquired surveillance video of the urban rail station area to obtain key point information of passengers in the video;

[0219] Based on the acquired passenger key point information, the key human motion features of each passenger are calculated according to the motion key feature calculation method.

[0220] Using a pre-trained random forest model, the key features of passengers' human body movements are classified to obtain the current behavior identification results of passengers.

[0221] Example 6

[0222] Embodiment 6 of the present invention provides an electronic device, including: a processor, a memory, and a computer program; wherein, the processor is connected to the memory, and the computer program is stored in the memory. When the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions for implementing a method for identifying abnormal passenger behavior in urban rail transit stations, the method including:

[0223] Obtain surveillance video of the urban rail station area;

[0224] Using a pre-trained detection model, passenger detection is performed on the acquired surveillance video of the urban rail station area to obtain key point information of passengers in the video;

[0225] Based on the acquired passenger key point information, the key human motion features of each passenger are calculated according to the motion key feature calculation method.

[0226] Using a pre-trained random forest model, the key features of passengers' human body movements are classified to obtain the current behavior identification results of passengers.

[0227] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0228] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0229] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0230] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment, whereby a series of operational steps are performed to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0231] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that, based on the technical solutions disclosed in the present invention, various modifications or variations that can be made by those skilled in the art without creative effort should be included within the scope of protection of the present invention.

Claims

1. A method for identifying abnormal passenger behavior in urban rail transit stations, characterized in that, include: Obtain surveillance video of the urban rail station area; Using a pre-trained detection model, passenger detection is performed on the acquired surveillance video of the urban rail station area to obtain key point information of passengers in the video; Based on the acquired passenger key point information, the key human motion features of each passenger are calculated according to the motion key feature calculation method. Using a pre-trained random forest model, the key features of passengers' human body movements are classified to obtain the current behavior identification results of passengers; Obtaining key point information of passengers in the video includes: extracting features from the input image using a VGG-19 network, and feeding the extracted feature maps as input to two convolutional neural networks; calculating the confidence and local affinity fields of human key points in the two convolutional neural networks to achieve key point localization and preliminary association; and using the Hungarian algorithm to solve for the final key point connection relationship based on the confidence and local affinity fields of human key points and the weights of the validity of the key point connection relationship. The specific formula for calculating the confidence level is as follows: Where, x i,j Let σ be the true coordinates of keypoint i of individual j in the image, σ be an empirical value, and p be the coordinates of a point in the image. The specific formula for calculating the local affinity field is as follows: Where, n c (p) represents the number of overlapping limbs c at point p; in, These are the coordinates of two key points that make up the limb c of the j-th individual.

2. The method for identifying abnormal passenger behavior in urban rail transit stations according to claim 1, characterized in that, The specific formula for calculating the weight of the validity of the key point connection relationship is as follows: in p is the key point A point on the line.

3. The method for identifying abnormal passenger behavior in urban rail transit stations according to claim 1, characterized in that, The specific calculation formula for the key features of the action is as follows: Where x i Let y be the x-coordinate of key point i. i Let L be the ordinate of key point i. i-j Let be the distance between keypoint i and keypoint j.

4. The method for identifying abnormal passenger behavior at urban rail stations according to claim 1, characterized in that, The classification process of the random forest model includes the following steps: For the training set, the bootstrap method is used to extract a new training set of size with replacement for each decision tree; for the m features in the training set, n features are randomly extracted at each non-leaf node of each tree, and the feature with the strongest classification ability is selected from these n features by evaluation indicators such as information gain ratio. Each decision tree is recursively constructed without pruning. All generated decision trees are combined into a random forest. The probability of the input data belonging to each category is obtained through the random forest, and the final predicted category is determined by voting.

5. A system for identifying abnormal passenger behavior in urban rail transit stations based on the method described in any one of claims 1-4, characterized in that, include: The acquisition module is used to acquire surveillance video of the urban rail station area; The detection module is used to detect passengers in the acquired surveillance video of the urban rail station area using a pre-trained detection model, and obtain key point information of passengers in the video. The calculation module is used to calculate the key human motion features of each passenger based on the acquired passenger key point information and according to the motion key feature calculation method. The classification module uses a pre-trained random forest model to classify the key features of passengers' human body movements and obtain the current behavior identification results of passengers.

6. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium is used to store computer instructions, which, when executed by a processor, implement the method for identifying abnormal passenger behavior at urban rail stations as described in any one of claims 1-4.

7. A computer program product, characterized in that, Includes a computer program, which, when run on one or more processors, is used to implement the method for identifying abnormal passenger behavior at urban rail stations as described in any one of claims 1-4.

8. An electronic device, characterized in that, include: The device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions for implementing the urban rail station passenger abnormal behavior identification method as described in any one of claims 1-4.

Images