Gate channel following behavior monitoring method, device and equipment and medium

Abstract

The invention discloses a gate channel trailing behavior monitoring method, device and equipment and a medium, relates to the technical field of image processing, and is applied to a server configured with a target trailing behavior recognition model. Comprising the steps of establishing target nodes corresponding to pedestrians in a video image corresponding to a to-be-recognized area and node identifiers corresponding to the target nodes, performing trajectory tracking on the pedestrians to obtain real-time trajectories, and storing the target nodes and the node identifiers in a preset relationship network; identifying the behavior state corresponding to each pedestrian, judging whether the behavior state meets a preset condition, and if yes, eliminating the corresponding node identifier in the preset relationship network to stop state identification of the pedestrian corresponding to the target node; and if not, determining that the current pedestrian has the following behavior, and transmitting the real-time trajectory of the pedestrian and the pedestrian image to a preset terminal device to start an emergency response measure for the following behavior. In this way, the accuracy of trailing behavior recognition can be improved.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a gate channel tailgating behavior monitoring method, device, equipment and medium. Background Art

[0002] At present, there are four methods used to determine whether the behavior of pedestrians crossing the turnstile is legal, namely: infrared anti-tailing method, face recognition anti-tailing method, traditional physical three-roller turnstile anti-tailing method and AB door anti-tailing method, but all of the above four methods have problems in some aspects.

[0003] First, the infrared anti-tailgating method is easily affected by the environment, is sensitive to obstacles, lacks effective notification means when tailing behavior is detected, and requires frequent maintenance and regular calibration. Secondly, the face recognition anti-tailgating method cannot accurately identify individuals under special circumstances, lacks effective notification means when detecting tailing behavior, has low efficiency in personnel passage, and is integrated inside the gate machine, making subsequent upgrades difficult, time-consuming, and costly. Then, the traditional physical three-roller gate anti-tailgating method has a slow passage speed, is inconvenient for pedestrians to carry luggage, cannot perform character recognition and tracking, has a high maintenance cost, and is not friendly to people with limited mobility. Finally, the AB door anti-tailgating method requires a large space and has low pedestrian passage efficiency. Summary of the invention

[0004] In view of this, the purpose of the present invention is to provide a gate channel tailgating behavior monitoring method, device, equipment and medium, which can improve the accuracy of tailgating behavior recognition, track the trajectory of pedestrians with tailgating behavior, and improve the efficiency of controlling pedestrians with tailgating behavior. The specific scheme is as follows:

[0005] In a first aspect, the present application provides a gate channel tailgating behavior monitoring method, which is applied to a server configured with a target tailgating behavior recognition model, wherein the target tailgating behavior recognition model is a model obtained by training based on an artificial intelligence model; wherein the method comprises:

[0006] Establish target nodes corresponding to each pedestrian in the video image corresponding to the area to be identified and node identifiers corresponding to the target nodes, track the trajectory of the pedestrian to obtain a real-time trajectory, and then save each target node and the corresponding node identifier to a preset relationship network; the preset relationship network is a relationship network established based on a preset graph algorithm;

[0007] Identify the behavior states corresponding to the pedestrians, and determine whether the obtained behavior states of the pedestrians meet the preset conditions. If the behavior states meet the preset conditions, eliminate the corresponding node identifiers in the preset relationship network to stop the state identification of the pedestrians corresponding to the corresponding target nodes.

[0008] If the behavior state does not meet the preset conditions, it is determined that the pedestrian is currently following, and the real-time trajectory and pedestrian image corresponding to the pedestrian are transmitted to a preset terminal device to initiate emergency response measures for the following behavior.

[0009] Optionally, the establishing of target nodes corresponding to the pedestrians in the video image corresponding to the area to be identified and node identifiers corresponding to the target nodes, and tracking the trajectories of the pedestrians to obtain real-time trajectories, includes:

[0010] Dividing the to-be-recognized area in the real-time to-be-recognized video by using a preset area division rule to obtain an to-be-processed area, and processing the to-be-processed video image corresponding to the to-be-processed area by using a preset resolution enhancement algorithm to obtain a target video image;

[0011] Target nodes corresponding to the pedestrians in the target video image and node identifiers corresponding to the target nodes are established by using several processes that can share information with each other, and the trajectories of the pedestrians are tracked to obtain real-time trajectories.

[0012] Optionally, the step of saving each of the target nodes and the corresponding node identifiers to a preset relationship network includes:

[0013] Based on the information shared between the processes, the target nodes corresponding to the pedestrians and the node identifiers corresponding to the target nodes are aligned in real time, so that the target nodes and the node identifiers corresponding to the target nodes that have been aligned in real time are stored in the preset relationship network.

[0014] Optionally, the tracking the pedestrian to obtain a real-time trajectory includes:

[0015] Acquire a real-time video to be identified, and determine a pedestrian position corresponding to each pedestrian in the area to be identified in the real-time video to be identified;

[0016] A real-time trajectory corresponding to the pedestrian is determined based on the position of each pedestrian.

[0017] Optionally, the identifying the behavior states corresponding to the pedestrians respectively and judging whether the obtained behavior states of the pedestrians meet preset conditions, if the behavior states meet the preset conditions, eliminating the corresponding node identifiers in the preset relationship network to stop the state identification of the pedestrians corresponding to the corresponding target nodes, comprises:

[0018] Identifying the behavior states respectively corresponding to the pedestrians to obtain the behavior states respectively corresponding to the pedestrians;

[0019] Determining whether the obtained behavior status of the pedestrians meets a preset condition, so as to obtain a determination result corresponding to each of the pedestrians;

[0020] Using a preset priority sorting algorithm and sorting the plurality of judgment results in descending order of priority to obtain a sorting result; wherein the priority corresponding to the judgment result representing that the preset condition is met is higher than the priority corresponding to the judgment result representing that the preset condition is not met;

[0021] Obtaining each of the judgment results from the sorting results in order from front to back;

[0022] If the judgment result obtained from the sorting result represents that the preset condition is met, the corresponding node identifier is eliminated in the preset relationship network to stop the state recognition operation of the pedestrian corresponding to the corresponding target node;

[0023] Correspondingly, if the behavior state does not meet the preset condition, determining that the pedestrian is currently following includes:

[0024] If the judgment result representation obtained from the sorting result does not satisfy the preset condition, it is determined that the current pedestrian is in a tailgating behavior.

[0025] Optionally, the identifying the behavior states corresponding to the pedestrians respectively and judging whether the obtained behavior states of the pedestrians meet preset conditions include:

[0026] The behavior status corresponding to each of the pedestrians is identified, and based on the behavior status, it is determined whether the behavior of the pedestrian crossing the gate is a tailgating behavior.

[0027] Optionally, the identifying the behavior status corresponding to each of the pedestrians, and judging whether the behavior of the pedestrian crossing the gate is a tailgating behavior based on the behavior status, includes:

[0028] Identify the behavior states corresponding to the pedestrians respectively, and determine whether the pedestrians are crossing the gate based on the behavior states;

[0029] If the pedestrian has not crossed the gate, jump to the step of identifying the behavior states corresponding to each of the pedestrians;

[0030] If the pedestrian is crossing the gate, an operation of determining whether the pedestrian's behavior is a tailgating behavior is triggered.

[0031] In a second aspect, the present application provides a gate channel tailgating behavior monitoring device, which is applied to a server configured with a target tailgating behavior recognition model, wherein the target tailgating behavior recognition model is a model obtained by training based on an artificial intelligence model; wherein the device comprises:

[0032] A trajectory tracking module is used to establish target nodes corresponding to each pedestrian in the video image corresponding to the area to be identified and node identifiers corresponding to the target nodes, and to track the pedestrians to obtain real-time trajectories, and then save the target nodes and corresponding node identifiers to a preset relationship network; the preset relationship network is a relationship network established based on a preset graph algorithm;

[0033] A state recognition module is used to recognize the behavior state corresponding to each of the pedestrians, and determine whether the obtained behavior state of the pedestrian meets the preset condition. If the behavior state meets the preset condition, the corresponding node identifier is eliminated in the preset relationship network to stop the state recognition of the pedestrian corresponding to the corresponding target node;

[0034] The trajectory transmission module is used to determine that the pedestrian is currently following if the behavior state does not meet the preset conditions, and transmit the real-time trajectory and pedestrian image corresponding to the pedestrian to a preset terminal device to initiate emergency response measures for the following behavior.

[0035] In a third aspect, the present application provides an electronic device, including:

[0036] Memory, used to store computer programs;

[0037] A processor is used to execute the computer program to implement the aforementioned gate channel tailgating behavior monitoring method.

[0038] In a fourth aspect, the present application provides a computer-readable storage medium for storing a computer program, wherein the computer program, when executed by a processor, implements the aforementioned gate channel tailgating behavior monitoring method.

[0039] As can be seen from the above, before performing tailgating behavior recognition, the present application needs to establish target nodes corresponding to each pedestrian in the video image corresponding to the area to be recognized and node identifiers corresponding to the target nodes, and track the trajectory of the pedestrian to obtain a real-time trajectory, and then save each target node and the corresponding node identifier to a preset relationship network; the preset relationship network is a relationship network established based on a preset graph algorithm; the behavior state corresponding to each pedestrian is identified, and it is determined whether the obtained behavior state of the pedestrian meets the preset conditions. If the behavior state meets the preset conditions, the corresponding node identifier is eliminated in the preset relationship network to stop the status recognition of the pedestrian corresponding to the corresponding target node; if the behavior state does not meet the preset conditions, it is determined that the current pedestrian has a tailing behavior, and the real-time trajectory corresponding to the pedestrian and the pedestrian image are transmitted to a preset terminal device so as to initiate emergency response measures for the tailing behavior.

[0040] It can be seen that the present application establishes target nodes and node identifiers corresponding to each pedestrian in the video image corresponding to the area to be identified, and tracks the pedestrian to obtain a real-time trajectory, and then saves each target node and the corresponding node identifier to a preset relationship network to identify the state corresponding to each pedestrian, so as to judge whether the pedestrian meets the preset legal conditions based on the obtained several behavior states, and if the behavior state meets the preset legal conditions, the node identifier is eliminated in the preset relationship network to stop the state identification of the pedestrian corresponding to the target node corresponding to the node identifier; if the behavior state corresponding to the pedestrian does not meet the preset legal conditions, the real-time trajectory corresponding to the pedestrian and the pedestrian image are transmitted to the preset terminal device so that the relevant personnel can control the pedestrian. In this way, the accuracy of the identification of the trailing behavior is improved, and the trajectory of the pedestrian with the trailing behavior is tracked, which improves the efficiency of controlling the pedestrian with the trailing behavior, thereby improving the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0042] Figure 1 A flow chart of a gate channel tailgating behavior monitoring method disclosed in the present application;

[0043] Figure 2 A schematic diagram of a flow chart of dividing and operating a region to be identified disclosed in the present application;

[0044] Figure 3 A schematic diagram of a specific overall architecture for pedestrian recognition disclosed in this application;

[0045] Figure 4 A schematic diagram of a specific flow chart of performing corresponding operations based on whether a pedestrian has crossed a gate disclosed in this application;

[0046] Figure 5 A schematic diagram of a specific result of analyzing pedestrian behavior disclosed in this application;

[0047] Figure 6 A flowchart of pedestrian behavior detection disclosed in this application;

[0048] Figure 7 A specific flowchart of pedestrian behavior detection disclosed in this application;

[0049] Figure 8 A schematic diagram of a result of identifying sample data disclosed in this application;

[0050] Fig. 9 A schematic diagram of identifying pedestrians in different situations disclosed in this application;

[0051] Fig.10 A specific schematic diagram of tailgating behavior disclosed in this application;

[0052] Fig.11 A specific flow chart of gate channel tailgating behavior monitoring disclosed in this application;

[0053] Fig.12 This is a schematic diagram of the structure of a gate channel tailgating behavior monitoring device disclosed in this application;

[0054] Fig.13 This is a structural diagram of an electronic device disclosed in this application. DETAILED DESCRIPTION

[0055] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] At present, among the methods used to determine whether the behavior of pedestrians crossing the gate is legal, the infrared anti-tailing method is easily affected by the environment, sensitive to obstacles, lacks effective notification means when detecting the existence of tailing behavior, and requires frequent maintenance. The face recognition anti-tailing method cannot accurately identify individuals under special circumstances, lacks effective notification means when detecting tailing behavior, has low efficiency of personnel passing, and is integrated inside the gate. It is time-consuming and costly to perform subsequent upgrades. The traditional physical three-roller gate anti-tailing method has a slow passage speed, is inconvenient for pedestrians to carry luggage, cannot perform character recognition and tracking, has high maintenance costs, and is not friendly to people with limited mobility. The AB door anti-tailing method has a large space requirement and low pedestrian passing efficiency. To this end, the present application provides a gate channel tailing behavior monitoring method that can improve the accuracy of tailing behavior recognition, track the trajectory of pedestrians with tailing behavior, and improve the efficiency of controlling pedestrians with tailing behavior.

[0057] See also Figure 1 As shown, an embodiment of the present invention discloses a gate channel tailgating behavior monitoring method, which is applied to a server configured with a target tailgating behavior recognition model, wherein the target tailgating behavior recognition model is a model obtained by training based on an artificial intelligence model; wherein the method comprises:

[0058] Step S11, establish target nodes corresponding to each pedestrian in the video image corresponding to the area to be identified and node identifiers corresponding to the target nodes, track the trajectory of the pedestrian to obtain a real-time trajectory, and then save each target node and the corresponding node identifier to a preset relationship network; the preset relationship network is a relationship network established based on a preset graph algorithm.

[0059] At present, in areas with high traffic volume such as railway stations or subway stations, the flow of pedestrians is relatively dense, which makes it difficult to identify the corresponding behavioral states of pedestrians. That is, in scenarios with dense traffic volume and high resolution, the existing algorithms have high difficulty in tracking the target trajectory due to the large mobility of people and the large number of targets in front of the gate, which is prone to misjudgment, resulting in insufficient recognition accuracy in local areas.

[0060] In this embodiment, before establishing the target nodes corresponding to the pedestrians in the video image corresponding to the area to be identified and the node identifiers corresponding to the target nodes, the embodiment of the present application needs to divide the area to be identified in the real-time video to be identified, and use a preset resolution enhancement algorithm to process the video images to be processed corresponding to the obtained several areas to be processed. Specifically, establishing the target nodes corresponding to the pedestrians in the video image corresponding to the area to be identified and the node identifiers corresponding to the target nodes, and tracking the trajectories of the pedestrians to obtain real-time trajectories, can include: dividing the area to be identified in the real-time video to be identified using a preset area division rule to obtain the area to be processed, and using a preset resolution enhancement algorithm to process the video image to be processed corresponding to the area to be processed to obtain the target video image; using several processes that can share information with each other to establish the target nodes corresponding to the pedestrians in the target video image and the node identifiers corresponding to the target nodes, and tracking the trajectories of the pedestrians to obtain real-time trajectories.

[0061] In the process of tracking the trajectory of pedestrians to obtain real-time trajectories, it is first necessary to determine the pedestrian positions corresponding to each pedestrian in the area to be identified, and then determine the real-time trajectory corresponding to the pedestrian based on the pedestrian positions. Specifically, tracking the trajectory of pedestrians to obtain real-time trajectories may include: obtaining real-time video to be identified, and determining the pedestrian positions corresponding to each pedestrian in the area to be identified in the real-time video to be identified; and determining the real-time trajectory corresponding to the pedestrian based on each pedestrian position.

[0062] In this embodiment, in order to improve the accuracy of the AI ​​(Artificial Intelligence) tailgating behavior detection model in detecting pedestrian behavior, Figure 2 As shown, Figure 2 The figure is a flowchart of the division and operation of the identification area. That is, in order to enable the AI ​​tailing behavior detection model to perform high-precision target detection on the gate-related area, the embodiment of the present application separately divides the key areas of concern, and uses a preset super-resolution algorithm and other technical means to improve the resolution of the divided areas, thereby improving the detection accuracy of pedestrian behavior. At the same time, in order to ensure the real-time feedback of information, for different detection areas, the embodiment of the present application enables separate processes for independent tracking, and at the same time, through shared memory technology, ensures that different processes can quickly obtain each other's detection results, so that the detection results are aligned in real time, further improving the accuracy of pedestrian behavior detection in the image.

[0063] In order to organize the target nodes corresponding to the pedestrians in the target video image established by each process and the node identifiers corresponding to the target nodes, the embodiment of the present application needs to share the information obtained after the execution of several processes that can share information with each other. Specifically, saving each target node and the corresponding node identifier to the preset relationship network can include: based on the information shared between processes, the target nodes corresponding to each pedestrian and the node identifiers corresponding to the target nodes are aligned in real time, so that the real-time aligned target nodes and the node identifiers corresponding to the target nodes are stored in the preset relationship network.

[0064] Step S12: identify the behavior status corresponding to each of the pedestrians, and determine whether the obtained behavior status of the pedestrians meets the preset conditions. If the behavior status meets the preset conditions, eliminate the corresponding node identifier in the preset relationship network to stop status identification of the pedestrian corresponding to the corresponding target node.

[0065] In order to speed up the management personnel's control over pedestrians based on tailgating behavior, after the AI ​​tailing behavior detection model detects and analyzes the trajectories of all targets in the image, it is necessary to re-sort all current trajectory results so that the results with abnormal behavior can be transmitted to the terminal in a timely manner and an alarm message can be generated for warning; if there is no abnormal behavior, the next detection process will be entered.

[0066] In this embodiment, the pedestrian state of the pedestrian is identified by using several independent processes that can share information in the target following behavior recognition model to obtain the behavior state corresponding to the pedestrian, and then the several behavior states are sorted in order of priority from high to low using a preset priority sorting algorithm. Specifically, the behavior state corresponding to each pedestrian is identified, and it is determined whether the obtained behavior state of the pedestrian meets the preset condition. If the behavior state meets the preset condition, the corresponding node identifier is eliminated in the preset relationship network to stop the state identification of the pedestrian corresponding to the corresponding target node, which may include: identifying the behavior state corresponding to each pedestrian to obtain the behavior state corresponding to each pedestrian; determining whether the obtained behavior state of the pedestrian meets the preset condition to obtain the judgment result corresponding to each pedestrian; using the preset priority sorting algorithm and sorting the several judgment results in order from high to low priority to obtain the sorting result; wherein, the priority corresponding to the judgment result representing the satisfaction of the preset condition is higher than the priority corresponding to the judgment result representing the non-satisfaction of the preset condition; and obtaining each judgment result from the sorting result in order from front to back.

[0067] In order to make high-precision judgments on the status of pedestrians, the target detection and target tracking algorithms are introduced in the embodiment of the present application. At the same time, targeted modeling is performed based on the characteristics of high-speed railway stations with dense flow of people. The overall architecture for pedestrian recognition is as follows: Figure 3 shown.

[0068] The embodiment of the present application utilizes an AI tailing behavior detection model to perform high-precision detection on the target in the identification area, and judges the behavior information corresponding to the current target: if the target has passed the gate and the behavior is legal, the target node corresponding to the target is eliminated in the preset relationship graph, and the id corresponding to the current target is archived; if the target has not passed the gate, the current id is included in the preset relationship network as an activated node, and the trajectory of this target is continuously tracked, and the target trajectory is predicted, and then the obtained real-time trajectory is stored in the database. Specifically, if the judgment result representation obtained from the sorting result satisfies the preset conditions, the corresponding node identifier is eliminated in the preset relationship network to stop the state identification operation of the pedestrian corresponding to the corresponding target node; accordingly, if the behavior state does not meet the preset conditions, it is determined that the current pedestrian has a tailing behavior, which may include: if the judgment result representation obtained from the sorting result does not meet the preset conditions, it is determined that the current pedestrian has a tailing behavior.

[0069] In this embodiment, after identifying the behavior state corresponding to the pedestrian, it is necessary to determine whether the obtained behavior state meets the preset conditions. Specifically, identifying the behavior state corresponding to each pedestrian and determining whether the obtained behavior state of the pedestrian meets the preset conditions may include: identifying the behavior state corresponding to each pedestrian and determining whether the behavior of the pedestrian crossing the gate is a tailgating behavior based on the behavior state. Figure 4 As shown, Figure 4 The flowchart is a flowchart of performing corresponding operations based on whether the pedestrian has crossed the gate. When the AI ​​tailing behavior detection model has completed the positioning of all targets in the current image, it is necessary to judge the state of the located target: if the target position has crossed the gate, the node identifier corresponding to the current target is eliminated, so that the state of the pedestrian corresponding to the target node corresponding to the node identifier is no longer identified in the subsequent process to reduce the misjudgment rate; if the target position has not crossed the gate, the trajectory corresponding to the current target node is stored in the database and continuously tracked, and the target node is added to the preset relationship diagram. Specifically, the behavior state corresponding to each pedestrian is identified, and whether the behavior of the pedestrian crossing the gate is a tailing behavior based on the behavior state can include: identifying the behavior state corresponding to each pedestrian, and judging whether the pedestrian is crossing the gate based on each behavior state; if the pedestrian has not crossed the gate, jumping to the step of identifying the behavior state corresponding to each pedestrian; if the pedestrian is crossing the gate, triggering the operation of judging whether the pedestrian's behavior is a tailing behavior.

[0070] In order to improve the accuracy of state judgment and trajectory tracking, the embodiment of the present application chooses to use a preset graph algorithm to construct a preset relationship graph, so that the AI ​​tailing behavior detection model can make a better judgment on whether the target has crossed the gate or whether there is any illegal tracking behavior.

[0071] Step S13: If the behavior state does not meet the preset conditions, it is determined that the pedestrian is currently following, and the real-time trajectory and pedestrian image corresponding to the pedestrian are transmitted to a preset terminal device to initiate emergency response measures for the following behavior.

[0072] In this embodiment, the present application embodiment chooses to deploy the computer vision algorithm to the native gate system of the high-speed railway station to identify and warn pedestrians who violate the rules and break into the gate, and assist on-site staff in efficiently managing entrances and exits. A schematic diagram of the result of analyzing pedestrian behavior is shown in FIG. Figure 5 As shown in the figure, the pedestrian in the black thick line frame in the upper right corner and with invasion displayed on the top of the head is the pedestrian who is following.

[0073] In this embodiment, the process of detecting pedestrian behavior is as follows: Figure 6 As shown: In places where security and anti-terrorism needs need to be checked, cameras are deployed to collect pedestrian images in the scene, and the collected pedestrian images are pre-processed. The pre-processed pedestrian images are uploaded to a server equipped with an AI tailing detection model, and the AI ​​tailing detection model identifies and processes the pedestrian images. When the AI ​​tailing detection model finds that the pedestrian in the pedestrian image is tailing, it will send a signal to the alarm, mobile phone, tablet, duty room computer and other terminal devices to notify the on-site management personnel to reach the current location of the tailing person in time based on the signal received by the terminal device, and control the tailing person. At the same time, the AI ​​tailing detection model will also start tracking the trajectory of the tailing person immediately after identifying the tailing behavior, and record the entire movement trajectory of the tailing person through image guidance to help the management personnel control the tailing person.

[0074] The specific process of pedestrian behavior detection is as follows: Figure 7As shown: When the camera is deployed at the railway station, it will collect images from the entrance gate to the security checkpoint, and then upload the images to the server. The AI ​​tailing detection model installed in the server will detect pedestrians in the image. When someone follows a person who has passed the gate authentication to enter the station, the AI ​​tailing detection model installed in the server will detect this behavior, mark the target information, and issue instructions, so that the alarm on the gate where the tailing behavior occurs will sound an alarm, and at the same time send an alarm message to the on-site management personnel on site, and display the image of the person who has tailed on the mobile phone or tablet computer to assist the on-site management personnel to identify the tailing person and control him.

[0075] In this embodiment, when multiple tailgating incidents occur simultaneously, the on-site management personnel will be unable to stop them in time. The AI ​​trajectory tracking function in the AI ​​tailgating detection model will record the target's movement trajectory and provide the target's current position in real time while marking the target, so that the on-site management personnel can control the tailing personnel according to the real-time trajectory prompts on the terminal. In a specific embodiment, the target movement trajectory can be that the target has currently boarded the escalator, left the escalator and walked to the security check location, has passed the security check, and is sitting in the fifth chair in the third row on the left side of the A1B1 waiting area.

[0076] In order to complete the high-precision detection of pedestrian behavior information and the tracking of pedestrian trajectories, the embodiment of the present application selected 102 targets in the sample video, and annotated 20 samples for each target. The AI ​​tailing behavior detection model was used to detect the behavior of each target, thereby obtaining 102 trajectories. Among them, 92 trajectories corresponded to targets that were compliant pedestrian targets, and 10 trajectories corresponded to targets that were not compliant pedestrian targets. The specific results are as follows: Figure 8 shown.

[0077] Among the non-compliant pedestrian targets, 6 groups were simulated by staff and 4 groups were real pedestrians. After being trained with 2,100 samples, the AI ​​tailing behavior detection model was tested in real scenes. When detecting 10 non-compliant pedestrian targets, 7 groups of non-compliant pedestrian targets were identified and detected. The specific results are as follows: Fig. 9 It is worth mentioning that the schematic diagram of 3 non-compliant pedestrian targets among the 10 non-compliant pedestrian targets is as follows Fig.10 shown.

[0078] In addition, the AI ​​stalking behavior detection model includes a video input module, a self-built model target detection module, a detection result filtering module, a target allocation module, a target tracking module, an object recognition module, and a data output module, and the schematic diagram is as follows: Fig.11 As shown:

[0079] First, the function of the video input module is to perform frame extraction on the video to be identified and generate a timestamp using a timestamp generator.

[0080] Secondly, the functions of the self-built model target detection module are to perform image preprocessing (ImagePreprocessing) on ​​the video frame, extract features from the video frame using the feature extraction network (Feature Extraction Network), region proposal network (RPN - Region Proposal Network), bounding box regression and classification (Bounding Box Regression & Classification) and detection result filtering (Detection Result Filtering). Among them, image preprocessing of video frames includes resizing (Resize), normalization (Normalization), data augmentation (Data Augmentation); feature extraction of video frames using the feature extraction network includes: convolutional layers (Convolutional Layers), residual blocks (Residual Blocks) and feature fusion module (Feature Fusion Module); region proposal network includes sliding window generator (Sliding Window Generator), anchor box regressor (Anchor Box Regressor) and classifier (Classifier); bounding box regression and classification include bounding box regression (Bounding Box Regression), Softmax classification (Softmax Classification) and confidence scoring (Confidence Scoring). Detection result filtering includes non-maximum suppression (NMS - Non-MaximumSuppression) and filter (Filter).

[0081] Furthermore, the target assignment module includes a Hungarian Algorithm and a distance measurement.

[0082] Then, the target tracking module includes Kalman filter, state prediction, observation update and error covariance calculation.

[0083] Subsequently, the object recognition module includes feature extraction, classifier, and object-human association.

[0084] Finally, the data output module includes data storage, real-time display, and behavior analysis interface.

[0085] As can be seen from the above, before the embodiment of the present application performs tailgating behavior monitoring in the gate channel, it is necessary to establish target nodes corresponding to each pedestrian in the video image corresponding to the area to be identified and node identifiers corresponding to the target nodes, and track the pedestrians to obtain real-time trajectories, and then save each target node and the corresponding node identifier to a preset relationship network to identify the states corresponding to each pedestrian, so as to determine whether the pedestrian meets the preset legal conditions based on the obtained several behavior states, and if the behavior state meets the preset legal conditions, then eliminate the node identifier in the preset relationship network to stop the state identification of the pedestrian corresponding to the target node corresponding to the node identifier; if the behavior state corresponding to the pedestrian does not meet the preset legal conditions, then transmit the real-time trajectory corresponding to the pedestrian and the pedestrian image to the preset terminal device so that relevant personnel can control the pedestrian. In this way, the accuracy of tailgating behavior identification is improved, and the trajectory of pedestrians with tailing behavior is tracked, which improves the efficiency of controlling pedestrians with tailing behavior.

[0086] Accordingly, see Fig.12 As shown, the present application also provides a gate channel tailgating behavior monitoring device, which is applied to a server configured with a target tailgating behavior recognition model, wherein the target tailgating behavior recognition model is a model obtained by training based on an artificial intelligence model; wherein the device comprises:

[0087] The trajectory tracking module 11 is used to establish target nodes corresponding to each pedestrian in the video image corresponding to the area to be identified and node identifiers corresponding to the target nodes, and to track the pedestrians to obtain real-time trajectories, and then save the target nodes and corresponding node identifiers to a preset relationship network; the preset relationship network is a relationship network established based on a preset graph algorithm;

[0088] The state recognition module 12 is used to recognize the behavior state corresponding to each of the pedestrians, and determine whether the obtained behavior state of the pedestrian meets the preset condition. If the behavior state meets the preset condition, the corresponding node identifier is eliminated in the preset relationship network to stop the state recognition of the pedestrian corresponding to the corresponding target node;

[0089] The trajectory transmission module 13 is used to determine that the pedestrian is currently following if the behavior state does not meet the preset conditions, and transmit the real-time trajectory and pedestrian image corresponding to the pedestrian to a preset terminal device to initiate emergency response measures for the following behavior.

[0090] As can be seen from the above, before the embodiment of the present application performs tailgating behavior monitoring in the gate channel, it is necessary to establish target nodes corresponding to each pedestrian in the video image corresponding to the area to be identified and node identifiers corresponding to the target nodes, and track the trajectory of the pedestrian to obtain a real-time trajectory, and then save each target node and the corresponding node identifier to a preset relationship network; the preset relationship network is a relationship network established based on a preset graph algorithm; the behavior state corresponding to each pedestrian is identified, and it is determined whether the obtained behavior state of the pedestrian meets the preset conditions. If the behavior state meets the preset conditions, the corresponding node identifier is eliminated in the preset relationship network to stop the status identification of the pedestrian corresponding to the corresponding target node; if the behavior state does not meet the preset conditions, it is determined that the current pedestrian has a tailgating behavior, and the real-time trajectory corresponding to the pedestrian and the pedestrian image are transmitted to a preset terminal device so as to initiate emergency response measures for the tailing behavior.

[0091] It can be seen that the embodiment of the present application establishes target nodes corresponding to each pedestrian in the video image corresponding to the area to be identified and node identifiers corresponding to the target nodes, and tracks the pedestrians to obtain real-time trajectories, and then saves each target node and the corresponding node identifier to a preset relationship network to identify the states corresponding to each pedestrian, so as to determine whether the pedestrian meets the preset legal conditions based on the obtained several behavior states, and if the behavior state meets the preset legal conditions, the node identifier is eliminated in the preset relationship network to stop the state identification of the pedestrian corresponding to the target node corresponding to the node identifier; if the behavior state corresponding to the pedestrian does not meet the preset legal conditions, the real-time trajectory corresponding to the pedestrian and the pedestrian image are transmitted to the preset terminal device so that the relevant personnel can control the pedestrian. In this way, the accuracy of the identification of the trailing behavior is improved, and the trajectory of the pedestrian with the trailing behavior is tracked, which improves the efficiency of controlling the pedestrian with the trailing behavior.

[0092] In some specific implementations, the state recognition module 11 may specifically include:

[0093] A region division unit is used to divide the to-be-identified region in the real-time to-be-identified video by using a preset region division rule to obtain a to-be-processed region, and to process the to-be-processed video image corresponding to the to-be-processed region by using a preset resolution enhancement algorithm to obtain a target video image;

[0094] The trajectory tracking unit is used to establish target nodes corresponding to each pedestrian in the target video image and node identifiers corresponding to the target nodes by using several processes that can share information with each other, and to track the trajectory of the pedestrian to obtain a real-time trajectory.

[0095] In some specific implementations, the state recognition module 11 may specifically include:

[0096] An information alignment unit is used to align the target nodes corresponding to each of the pedestrians and the node identifiers corresponding to the target nodes in real time based on the information shared between the processes, so as to store the target nodes and the node identifiers corresponding to the target nodes that have been aligned in real time into the preset relationship network.

[0097] In some specific implementations, the state recognition module 11 may specifically include:

[0098] A video acquisition unit, used to acquire a real-time video to be identified, and determine a pedestrian position corresponding to each pedestrian in the area to be identified in the real-time video to be identified;

[0099] A real-time trajectory determination unit is used to determine the real-time trajectory corresponding to the pedestrian based on the position of each pedestrian.

[0100] In some specific implementations, the state recognition module 12 may specifically include:

[0101] A state recognition unit, used to recognize the behavior states corresponding to the pedestrians, so as to obtain the behavior states corresponding to the pedestrians;

[0102] A condition judgment unit, used to judge whether the obtained behavior state of the pedestrians meets a preset condition, so as to obtain a judgment result corresponding to each of the pedestrians;

[0103] A result sorting unit, used to sort the plurality of judgment results in descending order of priority using a preset priority sorting algorithm to obtain a sorting result; wherein the priority corresponding to the judgment result representing that the preset condition is met is higher than the priority corresponding to the judgment result representing that the preset condition is not met;

[0104] A judgment result acquisition unit, used for acquiring each of the judgment results from the sorting results in order from front to back;

[0105] an identification eliminating unit, configured to eliminate the corresponding node identification in the preset relationship network if the judgment result representation obtained from the sorting result satisfies the preset condition, so as to stop the state identification operation of the pedestrian corresponding to the corresponding target node;

[0106] Accordingly, in some specific implementations, the state recognition module 12 may specifically include:

[0107] The first behavior judgment unit is used to determine that the current pedestrian has a trailing behavior if the judgment result representation obtained from the sorting result does not meet the preset condition.

[0108] In some specific implementations, the state recognition module 12 may specifically include:

[0109] The second behavior judgment unit is used to identify the behavior status corresponding to each of the pedestrians, and judge whether the behavior of the pedestrian crossing the gate is a tailgating behavior based on the behavior status.

[0110] In some specific implementations, the state recognition module 12 may specifically include:

[0111] a third behavior judgment unit, configured to identify the behavior states corresponding to the pedestrians, and judge whether the pedestrians are crossing the gate based on the behavior states;

[0112] A step jump unit, used for jumping to the step of identifying the behavior states corresponding to each of the pedestrians if the pedestrian has not crossed the gate;

[0113] The fourth behavior judgment unit is used to trigger an operation of judging whether the behavior of the pedestrian is a tailgating behavior if the pedestrian is crossing the gate.

[0114] Furthermore, the present application also discloses an electronic device. Fig.13 It is a structural diagram of an electronic device 20 shown according to an exemplary embodiment, and the content in the figure cannot be regarded as any limitation on the scope of use of this application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input and output interface 25 and a communication bus 26. Among them, the memory 22 is used to store a computer program, and the computer program is loaded and executed by the processor 21 to implement the relevant steps in the gate channel tailgating behavior monitoring method disclosed in any of the aforementioned embodiments. In addition, the electronic device 20 in this embodiment can specifically be an electronic computer.

[0115] In this embodiment, the power supply 23 is used to provide working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of the present application, and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs and is not specifically limited here.

[0116] In addition, the memory 22, as a carrier for storing resources, can be a read-only memory, a random access memory, a disk or an optical disk, etc. The resources stored thereon can include an operating system 221, a computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0117] The operating system 221 is used to manage and control the hardware devices and computer program 222 on the electronic device 20, which can be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program that can be used to complete the gate channel tailgating behavior monitoring method performed by the electronic device 20 disclosed in any of the aforementioned embodiments, the computer program 222 can further include a computer program that can be used to complete other specific tasks.

[0118] Furthermore, the present application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the aforementioned disclosed gate channel tailgating behavior monitoring method is implemented. The specific steps of the method can refer to the corresponding contents disclosed in the aforementioned embodiments, and will not be repeated here.

[0119] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0120] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0121] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0122] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0123] The technical solution provided by the present application is introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for general technicians in this field, according to the idea of ​​the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for monitoring tailgating behavior in a gate channel, characterized in that: Applied to a server configured with a target following behavior recognition model, wherein the target following behavior recognition model is a model obtained by training based on an artificial intelligence model; wherein the method comprises: Establish target nodes corresponding to each pedestrian in the video image corresponding to the area to be identified and node identifiers corresponding to the target nodes, track the trajectory of the pedestrian to obtain a real-time trajectory, and then save each target node and the corresponding node identifier to a preset relationship network; the preset relationship network is a relationship network established based on a preset graph algorithm; Identify the behavior states corresponding to the pedestrians, and determine whether the obtained behavior states of the pedestrians meet the preset conditions. If the behavior states meet the preset conditions, eliminate the corresponding node identifiers in the preset relationship network to stop the state identification of the pedestrians corresponding to the corresponding target nodes. If the behavior state does not meet the preset conditions, it is determined that the pedestrian is currently following, and the real-time trajectory and pedestrian image corresponding to the pedestrian are transmitted to a preset terminal device to initiate emergency response measures for the following behavior.

2. The gate channel tailgating behavior monitoring method according to claim 1 is characterized in that: The step of establishing target nodes corresponding to the pedestrians in the video image corresponding to the area to be identified and node identifiers corresponding to the target nodes, and tracking the trajectories of the pedestrians to obtain real-time trajectories includes: Dividing the to-be-recognized area in the real-time to-be-recognized video by using a preset area division rule to obtain an to-be-processed area, and processing the to-be-processed video image corresponding to the to-be-processed area by using a preset resolution enhancement algorithm to obtain a target video image; Target nodes corresponding to the pedestrians in the target video image and node identifiers corresponding to the target nodes are established by using several processes that can share information with each other, and the trajectories of the pedestrians are tracked to obtain real-time trajectories.

3. The gate channel tailgating behavior monitoring method according to claim 2 is characterized in that: The step of saving each of the target nodes and the corresponding node identifiers to a preset relationship network includes: Based on the information shared between the processes, the target nodes corresponding to the pedestrians and the node identifiers corresponding to the target nodes are aligned in real time, so that the target nodes and the node identifiers corresponding to the target nodes that have been aligned in real time are stored in the preset relationship network.

4. The gate channel tailgating behavior monitoring method according to claim 1 is characterized in that: The step of tracking the pedestrian to obtain a real-time trajectory includes: Acquire a real-time video to be identified, and determine a pedestrian position corresponding to each pedestrian in the area to be identified in the real-time video to be identified; A real-time trajectory corresponding to the pedestrian is determined based on the position of each pedestrian.

5. The gate channel tailgating behavior monitoring method according to claim 1 is characterized in that: The identifying the behavior states corresponding to the pedestrians respectively and judging whether the obtained behavior states of the pedestrians meet the preset conditions, if the behavior states meet the preset conditions, eliminating the corresponding node identifiers in the preset relationship network to stop the state identification of the pedestrians corresponding to the corresponding target nodes, comprises: Identifying the behavior states respectively corresponding to the pedestrians to obtain the behavior states respectively corresponding to the pedestrians; Determining whether the obtained behavior status of the pedestrians meets a preset condition, so as to obtain a determination result corresponding to each of the pedestrians; Using a preset priority sorting algorithm and sorting the plurality of judgment results in descending order of priority to obtain a sorting result; wherein the priority corresponding to the judgment result representing that the preset condition is met is higher than the priority corresponding to the judgment result representing that the preset condition is not met; Obtaining each of the judgment results from the sorting results in order from front to back; If the judgment result obtained from the sorting result represents that the preset condition is met, the corresponding node identifier is eliminated in the preset relationship network to stop the state recognition operation of the pedestrian corresponding to the corresponding target node; Correspondingly, if the behavior state does not meet the preset condition, determining that the pedestrian is currently following includes: If the judgment result representation obtained from the sorting result does not satisfy the preset condition, it is determined that the current pedestrian has a tailgating behavior.

6. The gate channel tailgating behavior monitoring method according to any one of claims 1 to 5, characterized in that: The identifying the behavior states corresponding to the pedestrians respectively and judging whether the obtained behavior states of the pedestrians meet preset conditions include: The behavior status corresponding to each of the pedestrians is identified, and based on the behavior status, it is determined whether the behavior of the pedestrian crossing the gate is a tailgating behavior.

7. The gate channel tailgating behavior monitoring method according to claim 6 is characterized in that: The identifying the behavior states corresponding to the pedestrians respectively, and judging whether the behavior of the pedestrians crossing the gate is a tailgating behavior based on the behavior states, includes: Identify the behavior states corresponding to the pedestrians respectively, and determine whether the pedestrians are crossing the gate based on the behavior states; If the pedestrian has not crossed the gate, jump to the step of identifying the behavior states corresponding to each of the pedestrians; If the pedestrian is crossing the gate, an operation of determining whether the pedestrian's behavior is a tailgating behavior is triggered.

8. A gate channel tailgating behavior monitoring device, characterized in that: Applied to a server configured with a target following behavior recognition model, wherein the target following behavior recognition model is a model obtained by training based on an artificial intelligence model; wherein the device comprises: A trajectory tracking module is used to establish target nodes corresponding to each pedestrian in the video image corresponding to the area to be identified and node identifiers corresponding to the target nodes, and to track the pedestrians to obtain real-time trajectories, and then save the target nodes and corresponding node identifiers to a preset relationship network; the preset relationship network is a relationship network established based on a preset graph algorithm; A state recognition module is used to recognize the behavior state corresponding to each of the pedestrians, and determine whether the obtained behavior state of the pedestrian meets the preset condition. If the behavior state meets the preset condition, the corresponding node identifier is eliminated in the preset relationship network to stop the state recognition of the pedestrian corresponding to the corresponding target node; The trajectory transmission module is used to determine that the pedestrian is currently following if the behavior state does not meet the preset conditions, and transmit the real-time trajectory and pedestrian image corresponding to the pedestrian to a preset terminal device to initiate emergency response measures for the following behavior.

9. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor is used to execute the computer program to implement the steps of the gate channel tailgating behavior monitoring method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: Used to store a computer program, wherein when the computer program is executed by a processor, the steps of the gate channel tailgating behavior monitoring method as described in any one of claims 1 to 7 are implemented.

Images