Intelligent park security risk perception system and method with prediction capability

Abstract

This invention discloses an intelligent park security risk perception system and method with predictive capabilities. The method includes: collecting facial feature data registered at the park's access control system; collecting video data from T monitoring areas in real time; extracting real-time facial feature data appearing in the video data of the t-th monitoring area; comparing the real-time facial feature data with the registered facial feature data, the comparison results including whether there is an entry record or not; marking the real-time facial feature data without an entry record as abnormal personnel features; pre-setting deliberate avoidance zones for each monitoring area; using a tracking algorithm to track abnormal personnel with abnormal personnel features, and recording the deliberate avoidance zones in q monitoring areas where the abnormal personnel features appear, marking them as passing avoidance zones; and connecting all the recorded passing avoidance zones sequentially to form the movement trajectory of the abnormal personnel; thereby improving the park's security risk perception capabilities.

Description

Technical Field

[0001] This invention relates to the field of park security management technology, specifically to an intelligent park security risk perception system and method with predictive capabilities. Background Technology

[0002] As smart parks continue to develop and expand in scale, security risk management within these parks becomes increasingly complex and important. Traditional security monitoring systems typically rely on single behavioral characteristics, such as personnel entry and exit records and simple abnormal behavior detection, to assess potential security risks. However, this approach has significant limitations, failing to comprehensively capture personnel behavior patterns, resulting in inaccurate identification of abnormal behavior and a high false alarm rate.

[0003] In existing technologies, various behavioral characteristics are often analyzed in isolation, failing to fully consider the complex and close relationships between them. This lack of in-depth analysis makes it difficult for security systems to effectively identify potential risks and provide timely warnings of possible security incidents. In addition, frequent false alarms not only increase the workload of security personnel, but may also lead to negative attitudes among managers, further weakening the effectiveness of security management.

[0004] How to deeply explore the correlation between various behavioral characteristics and comprehensively and accurately assess the behavior of abnormal personnel has become a technical problem that urgently needs to be solved in the field of smart park security management. Summary of the Invention

[0005] The purpose of this invention is to provide a smart campus security risk perception system and method with predictive capabilities to address the shortcomings of the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for predictive intelligent park security risk perception, comprising:

[0007] Collect facial feature data from people registering at park access control;

[0008] Real-time acquisition of video data from T monitored areas within the park;

[0009] Extract real-time facial feature data from the video data appearing in the t-th monitoring area, t∈T. Compare the real-time facial feature data with the registered facial feature data to obtain the comparison results. The comparison results include whether there is an entry record or not. Mark the real-time facial feature data without an entry record as abnormal personnel features.

[0010] Pre-set designated areas to be avoided in each monitoring zone;

[0011] Using a tracking algorithm, we track abnormal individuals with unusual characteristics and record the areas in which these characteristics appear in q monitored areas. These areas are marked as "passing-through avoidance areas". We then connect all the recorded passing-through avoidance areas to form the movement trajectory of the abnormal individuals, where q is less than or equal to T.

[0012] Image frames of people with abnormal behavior are captured from video data, input into the action recognition model, and the avoidance actions of the abnormal people are obtained. The number of avoidance actions per unit time is counted.

[0013] Collect the current light intensity and humidity, and combine them with the number of avoidance actions to generate an avoidance action coefficient;

[0014] Obtain movement characteristic data and height data of abnormal individuals based on video data;

[0015] The movement trajectory, avoidance action coefficient, movement characteristic data, and height data are input into the behavioral characteristic analysis model to obtain the behavioral analysis results;

[0016] If the behavior analysis results indicate abnormal behavior, an alert will be issued.

[0017] Furthermore, the avoidance actions include looking down and covering the face, and the number of times the head is looked down and the face is covered is counted per unit time; the avoidance action coefficient generation includes:

[0018] The avoidance action value is obtained by weighted summing of the number of times the head is lowered and the face is covered per unit time. The environmental impact value is obtained by weighted summing of the current light intensity and humidity. The avoidance action coefficient is obtained by dividing the avoidance action value by the environmental impact value.

[0019] Furthermore, methods for extracting movement characteristic data of abnormal individuals include:

[0020] Motion characteristic data includes the average movement speed and the standard deviation of movement speed;

[0021] The location coordinates of the abnormal personnel in continuous frames of video data are mapped to the electronic map of the park to obtain the continuous location coordinate sequence of the abnormal personnel within a unit time interval.

[0022] Calculate the speed S moving per unit interval:

[0023]

[0024] X i+1 X represents the x-coordinate of the (i+1)th position. i Y represents the x-coordinate of the i-th position. i+1 Y represents the ordinate of the (i+1)th position. i The ordinate represents the coordinate of the i-th position;

[0025] The calculated movement speeds are used to create a set of movement speeds, and the average movement speed and standard deviation of the movement speeds within the set are calculated.

[0026] Furthermore, methods for obtaining height data of individuals with abnormal heights include:

[0027] Using video processing techniques, the top and bottom pixel coordinates of reference objects and unusual persons are obtained in the video data image;

[0028] The height of the reference object in the video data image is obtained based on the top and bottom pixel coordinates of the reference object; similarly, the height of the abnormal person in the video data image is obtained based on the top and bottom pixel coordinates of the abnormal person; the ratio of the height of the reference object in the video data image to the actual height of the reference object is marked as the scaling factor; the height of the abnormal person in the video data image is multiplied by the scaling factor to obtain the height data of the abnormal person.

[0029] Furthermore, the training methods for behavioral feature analysis models include:

[0030] K sets of action trajectories, avoidance action coefficients, height of abnormal personnel, and movement feature data are collected in advance and combined into K sets of feature vectors, where K is an integer greater than 1. Supervised learning methods are used, such as support vector machines, random forests, or deep neural networks, to train the behavior feature analysis model. Normal and abnormal behaviors are pre-labeled on the feature vectors to form a dataset of known behaviors. The dataset is divided into training and testing sets for training and validation of the behavior feature analysis model, and the output behavior feature analysis model that meets the preset accuracy is obtained.

[0031] Furthermore, it also includes:

[0032] Collect wind force and direction data for each monitored area;

[0033] Based on wind force and direction data and light intensity, the system analyzes whether the video data of each monitoring area is abnormal. If abnormal, an activation command is generated; if normal, no activation command is generated, and the activation command is sent to the backup server and the data switching switch. The backup server can perform real-time analysis of the video data. The backup server is activated according to the activation command, and the data switching switch transmits the video data to the backup server for analysis according to the activation command.

[0034] Furthermore, methods for analyzing whether video data for each monitored area is abnormal based on wind force and direction data and light intensity include:

[0035] When the wind force level of the r-th monitoring area is greater than the preset wind force level, analyze the offset direction of the flexible ribbon in the video data of the r-th monitoring area to see if it is consistent with the wind direction. If it is inconsistent, generate an activation command; if it is consistent, do not generate an activation command.

[0036] If the wind force level is less than or equal to the preset wind force level, the real-time collected light intensity and the real-time image frames from the video data of the r-th monitoring area will be input into the monitoring anomaly analysis model to determine whether the image frames are abnormal. If the image frames are abnormal, an activation command will be generated; if the image frames are normal, an activation command will not be generated.

[0037] Furthermore, the method for analyzing the offset direction of the flexible ribbon in the video data of the r-th monitoring area includes:

[0038] In the video data display, a pre-defined planar coordinate system is used, with the fixed end of the flexible ribbon as the origin of the planar coordinate system. The coordinates of each pixel in the area occupied by the flexible ribbon in the display at the current moment are obtained. The quadrant of the planar coordinate system where the flexible ribbon is located is determined based on the coordinates of each pixel. The offset direction of the flexible ribbon is determined based on the pre-defined direction of the quadrant.

[0039] Furthermore, the training method for the monitoring anomaly analysis model includes:

[0040] E sets of training data are collected in advance. Each set of training data includes the illumination intensity and the corresponding image frame, where E is an integer greater than 1. The training data is used as the input to the monitoring anomaly analysis model. The monitoring anomaly analysis model takes the illumination intensity corresponding to each set of image frames as the output and the actual illumination intensity corresponding to each set of image frames as the prediction target. The training objective is to minimize the sum of prediction errors for all illumination intensities. The monitoring anomaly analysis model is trained until the sum of prediction errors converges, at which point training stops. The monitoring anomaly analysis model uses a convolutional neural network.

[0041] A predictive intelligent park security risk perception system, used in the aforementioned predictive intelligent park security risk perception method, includes:

[0042] The first data acquisition module is used to collect facial feature data for park access control registration.

[0043] The second acquisition module is used to acquire video data from T monitored areas in the park in real time.

[0044] The first analysis module extracts real-time facial feature data from the video data appearing in the t-th monitoring area, where t∈T. It compares the real-time facial feature data with the registered facial feature data to obtain the comparison results. The comparison results include whether there is an entry record or not. The real-time facial feature data without an entry record is marked as abnormal personnel features.

[0045] The settings module allows you to pre-set the areas to be deliberately avoided in each monitoring zone;

[0046] The second analysis module uses a tracking algorithm to track abnormal individuals with abnormal characteristics and records the areas deliberately avoided by these individuals in q monitoring areas. These areas are marked as "passing-through avoidance areas". All the recorded passing-through avoidance areas are connected sequentially to form the movement trajectory of the abnormal individuals, where q is less than or equal to T.

[0047] The third analysis module extracts image frames of people with abnormal behavior from the video data, inputs them into the action recognition model, obtains the avoidance actions of the abnormal people, and counts the number of avoidance actions per unit time.

[0048] The fourth analysis module collects the current light intensity and humidity, and combines this with the number of avoidance actions to generate an avoidance action coefficient.

[0049] The fifth analysis module is used to obtain the movement characteristics and height data of abnormal persons based on video data;

[0050] The comprehensive analysis module is used to input movement trajectory, avoidance action coefficient, movement characteristic data and height data into the behavior characteristic analysis model to obtain behavior analysis results;

[0051] The early warning module issues an early warning if the behavior analysis results indicate abnormal behavior.

[0052] The technical effects and advantages of the intelligent park security risk perception system and method with predictive capabilities provided by this invention are as follows:

[0053] 1. By inputting multiple behavioral characteristics into the behavioral characteristic analysis model, the behavior of abnormal personnel can be comprehensively and accurately assessed. Each characteristic reflects the behavioral patterns and potential risks of personnel from different perspectives, and their relationship with the behavioral analysis results is close and complex. By deeply analyzing these relationships, the model can more effectively identify abnormal behavior, improve the park's ability to perceive security risks, effectively reduce false alarms, and the highly accurate early warning not only improves the efficiency of security prevention, but also reduces the impact of park managers' passive negligence.

[0054] 2. This embodiment aims to prevent the video surveillance system from being hacked and to play pre-recorded videos, thereby avoiding the loss of the function of warning of abnormal behavior of abnormal personnel; by detecting the abnormality of video data, when an anomaly is detected, the system will activate the backup server to analyze the real-time collected video data, improve system security, and prevent intrusion attacks.

[0055] By collecting real-time wind force and direction data and light intensity, the system can sense whether the video data matches the actual environmental conditions. When the direction of the flexible ribbon in the video is inconsistent with the actual wind direction, or the light intensity does not match the image frame, the system judges that the video may have been tampered with or replaced, and generates an activation command in a timely manner. The multi-verification mechanism effectively prevents hackers from deceiving the monitoring system by playing pre-recorded videos, and enhances the system's ability to resist intrusion and tampering.

[0056] By detecting anomalies in video data, the system automatically activates a backup server for in-depth analysis when an anomaly is detected, through intelligent switching of the data switching switch. This ensures that monitoring and analysis can continue even in abnormal situations, guaranteeing that the monitoring system can continuously and accurately identify and warn of abnormal personnel. It also avoids monitoring blind spots caused by video replacement, ensuring the effective implementation of park security measures. Attached Figure Description

[0057] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0058] Figure 1 This is a schematic diagram of the intelligent park security risk perception system with predictive capabilities in Embodiment 1 of the present invention;

[0059] Figure 2 This is a schematic diagram of the intelligent park security risk perception system with predictive capabilities in Embodiment 2 of the present invention;

[0060] Figure 3 This is a flowchart of the intelligent park security risk perception method with predictive capabilities in Embodiment 3 of the present invention. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0062] It should be noted that when a component is said to be "fixed to" another component, it can be directly attached to the other component or there may be an intervening component. When a component is said to be "connected to" another component, it can be directly connected to the other component or there may be an intervening component.

[0063] Example 1

[0064] like Figure 1 As shown, the intelligent park security risk perception system and method with predictive capabilities in this embodiment includes a first acquisition module, a second acquisition module, a first analysis module, a setting module, a second analysis module, a third analysis module, a fourth analysis module, a fifth analysis module, and a comprehensive analysis module. Each module is connected via wired and / or wireless means.

[0065] The first data acquisition module is used to collect facial feature data of people registering at the park's access control system and store it in the registration database.

[0066] The second acquisition module is used to acquire video data from T monitored areas in the park in real time.

[0067] The first analysis module extracts real-time facial feature data from the video data appearing in the t-th monitoring area, where t∈T. It compares the real-time facial feature data with the registered facial feature data to obtain the comparison results. The comparison results include whether there is an entry record or not. The real-time facial feature data without an entry record is marked as abnormal personnel features.

[0068] Methods for extracting real-time facial feature data include:

[0069] Using personnel recognition software, the regions containing human faces are extracted from the video data frames of the t-th monitored area to obtain real-time facial feature data.

[0070] The settings module allows for the pre-setting of deliberate avoidance zones for each monitoring area. These deliberate avoidance zones are the edge areas or obstructed areas of the monitoring area's cameras. Staff can set these zones based on the specific environment of the monitoring area. The reason is that the sensitivity in obstructed or edge areas may be lower, allowing suspicious individuals to bypass the automatic alarm mechanism.

[0071] The second analysis module performs real-time analysis of video data from T monitoring areas. Using algorithms such as SORT and DeepSORT, it tracks abnormal individuals with unusual characteristics and records areas deliberately avoided by these individuals within q monitoring areas. These areas are marked as "passing-through avoidance areas." All recorded passing-through avoidance areas are then connected sequentially to form the movement trajectory of the abnormal individuals, where q is less than or equal to T.

[0072] Methods for obtaining the movement patterns of individuals with unusual behavior include:

[0073] The actual coordinates of all the areas to be avoided are mapped onto the electronic map of the park. All the areas to be avoided are then connected in sequence on the electronic map to form the movement trajectory of the abnormal personnel.

[0074] The third analysis module captures image frames of individuals with abnormal behavior from the video data, inputs them into the action recognition model, obtains the avoidance actions of the individuals with abnormal behavior, including looking down and covering their faces, and counts the number of times they look down and cover their faces per unit time.

[0075] The fourth analysis module collects the light intensity and humidity at the current moment, and combines this with the number of times the head is lowered and the face is covered per unit time to generate an avoidance action coefficient.

[0076] The generation of avoidance action coefficients includes:

[0077]

[0078] In the formula, All are preset weighting coefficients; XS represents the avoidance action coefficient, the larger the value, the more abnormal the person is; DT represents the number of times the head is lowered per unit time; ZD represents the number of times the face is covered per unit time; GZ represents the light intensity at the current moment; SD represents the humidity at the current moment. The stronger the light intensity, the more avoidance actions will occur when the person is blocking sunlight. Similarly, the higher the humidity value, the more avoidance actions will occur when the person is blocking rain or fog. In order to avoid the influence of light intensity and humidity, the light intensity and humidity are weighted and used as the denominator. The generated avoidance action coefficient can more realistically reflect the abnormality of the person, highlight the real abnormal behavior, and reduce the interference of other environmental factors.

[0079] The fifth analysis module is used to obtain the movement characteristic data and height data of abnormal persons based on video data. The movement characteristic data includes the average movement speed and the standard deviation of movement speed.

[0080] Methods for extracting movement characteristic data of abnormal individuals include:

[0081] The location coordinates of the abnormal personnel in continuous frames of video data are mapped to the electronic map of the park to obtain the continuous location coordinate sequence of the abnormal personnel within a unit interval time.

[0082] Calculate the speed S moving per unit interval:

[0083]

[0084] X i+1 X represents the x-coordinate of the (i+1)th position. i Y represents the x-coordinate of the i-th position. i+1 Y represents the ordinate of the (i+1)th position. i The ordinate represents the coordinate of the i-th position.

[0085] The calculated movement speeds are used to create a set of movement speeds, and the average movement speed and standard deviation of the movement speeds within the set are calculated.

[0086] Methods for obtaining height data of individuals with abnormal heights include:

[0087] Using video processing techniques, the top and bottom pixel coordinates of reference objects and unusual persons are obtained in the video data image;

[0088] The height of the reference object in the video data image is obtained based on the top and bottom pixel coordinates of the reference object; similarly, the height of the abnormal person in the video data image is obtained based on the top and bottom pixel coordinates of the abnormal person; the ratio of the height of the reference object in the video data image to the actual height of the reference object is marked as the scaling factor; the height of the abnormal person in the video data image is multiplied by the scaling factor to obtain the height data of the abnormal person.

[0089] The comprehensive analysis module is used to input movement trajectory, avoidance action coefficient, movement characteristic data and height data into the behavior characteristic analysis model to obtain behavior analysis results.

[0090] The early warning module will issue an early warning if the behavior analysis results indicate abnormal behavior, and will not issue an early warning if the behavior analysis results indicate normal behavior.

[0091] Movement trajectory refers to the movement path of an abnormal person within a monitored area, which consists of a series of spatiotemporal coordinate points.

[0092] The movement trajectory reflects the direction of movement of abnormal individuals, the areas they pass through, and whether they choose routes commonly used by the public; the avoidance action coefficient quantifies the degree to which abnormal individuals perform avoidance actions per unit of time, indicating that individuals intentionally avoid cameras, conceal their identities, and have potential illegal intentions; excessively fast movement, exceeding normal movement speed, indicates that individuals are eager to leave the scene or evade tracking; excessively slow movement, below normal movement speed, may indicate that individuals are observing the environment, searching for targets, or avoiding attention; speed changes: stable speed may indicate normal activity, while drastic changes indicate abnormal behavior; high standard deviation, large speed fluctuations, may indicate that individuals are avoiding monitoring, observing the environment, or looking for opportunities; low standard deviation, stable speed, may indicate that individuals are performing specific tasks or have clear objectives; secondly, the height of abnormal individuals is considered to reduce the impact of height on movement speed and improve the predictive accuracy of the behavioral characteristic analysis model.

[0093] Training methods for behavioral feature analysis models include:

[0094] K sets of action trajectories, avoidance action coefficients, height of abnormal personnel, and movement feature data are collected in advance and combined into K sets of feature vectors, where K is an integer greater than 1. Supervised learning methods, such as support vector machine (SVM), random forest, and deep neural network, are used to train the behavior feature analysis model. Normal and abnormal behaviors are pre-labeled for the feature vectors to form a dataset of known behaviors. The dataset is divided into training set and test set for training and validation of the behavior feature analysis model, and the output behavior feature analysis model that meets the preset accuracy is achieved.

[0095] By inputting various behavioral characteristics into the behavioral characteristic analysis model, the behavior of abnormal personnel can be comprehensively and accurately assessed. Each characteristic reflects the behavioral patterns and potential risks of personnel from different perspectives, and their relationship with the behavioral analysis results is close and complex. By deeply analyzing these relationships, the model can more effectively identify abnormal behavior, improve the park's ability to perceive security risks, effectively reduce false alarms, and the highly accurate early warning not only improves the efficiency of security prevention but also reduces the impact of park managers' passive negligence.

[0096] Example 2

[0097] Please see Figure 2 As shown, this embodiment provides an intelligent campus security risk perception system with predictive capabilities, and also includes:

[0098] The third acquisition module is used to collect wind force and direction data for each monitoring area, specifically through anemometers or ultrasonic anemometers installed in each monitoring area.

[0099] The monitoring anomaly analysis module analyzes the video data of each monitored area for anomalies based on wind force and direction data and light intensity. If an anomaly is found, an activation command is generated; otherwise, no activation command is generated, and the activation command is sent to the backup server and the data switching switch. The backup server can perform real-time analysis of the video data. The backup server is activated according to the activation command. The input terminal of the data switching switch is connected to the data output line of the video acquisition device, and the output terminal of the data switching switch is connected to the analysis server and the backup server respectively. When the data switching switch receives the activation command, it transmits the video data to the backup server for analysis according to the activation command.

[0100] Methods for analyzing whether video data in each monitored area is abnormal based on wind force and direction data and light intensity include:

[0101] When the wind force level of the r-th monitoring area is greater than the preset wind force level, analyze the offset direction of the flexible ribbon in the video data of the r-th monitoring area to see if it is consistent with the wind direction. If it is inconsistent, generate an activation command; if it is consistent, do not generate an activation command.

[0102] If the wind force level is less than or equal to the preset wind force level, it means there is no wind or the wind is weak, which will cause the flexible ribbon to have a limited displacement range and is prone to causing the false generation of the activation command. Therefore, when the wind force level is less than or equal to the preset wind force level, the real-time collected light intensity and the real-time image frames from the video data of the r-th monitoring area are input into the monitoring anomaly analysis model to determine whether the image frame is abnormal. If the image frame is abnormal, an activation command is generated; if the image frame is normal, no activation command is generated.

[0103] Methods for analyzing the offset direction of the flexible ribbon in the video data of the r-th monitoring area include:

[0104] In the video data display, a pre-defined planar coordinate system is used, with the fixed end of the flexible ribbon as the origin of the planar coordinate system. The coordinates of each pixel in the area occupied by the flexible ribbon in the display at the current moment are obtained. The quadrant of the planar coordinate system where the flexible ribbon is located is determined based on the coordinates of each pixel. The offset direction of the flexible ribbon is determined based on the pre-defined direction of the quadrant.

[0105] The first quadrant: both x and y are positive, located to the upper right of the origin.

[0106] The second quadrant: x is negative and y is positive, located to the upper left of the origin.

[0107] The third quadrant: both x and y are negative, located to the lower left of the origin.

[0108] The fourth quadrant: x is positive and y is negative, located to the lower right of the origin.

[0109] Methods for determining the quadrant of the planar coordinate system in which the flexible ribbon is located based on the coordinates of each pixel include:

[0110] The quadrant containing the most pixels is designated as the quadrant where the flexible ribbon resides.

[0111] Training methods for monitoring anomaly analysis models include:

[0112] E sets of training data are collected in advance. Each set of training data includes the illumination intensity and the corresponding image frame, where E is an integer greater than 1. The training data is used as the input to the monitoring anomaly analysis model. The monitoring anomaly analysis model takes the illumination intensity corresponding to each set of image frames as the output and the actual illumination intensity corresponding to each set of image frames as the prediction target. The training objective is to minimize the sum of prediction errors for all illumination intensities. The monitoring anomaly analysis model is trained until the sum of prediction errors converges, at which point training stops. The monitoring anomaly analysis model uses a deep learning model (such as a convolutional neural network) to predict illumination intensity from image frames.

[0113] This embodiment aims to prevent the video surveillance system from being hacked and to play pre-recorded videos, thereby avoiding the loss of the function of warning of abnormal behavior of abnormal personnel; by detecting the abnormality of video data, when an anomaly is detected, the system will activate a backup server to analyze the real-time collected video data, thereby improving system security and preventing intrusion attacks.

[0114] By collecting real-time wind force and direction data and light intensity, the system can sense whether the video data matches the actual environmental conditions. When the direction of the flexible ribbon in the video is inconsistent with the actual wind direction, or the light intensity does not match the image frame, the system judges that the video may have been tampered with or replaced, and generates an activation command in a timely manner. The multi-verification mechanism effectively prevents hackers from deceiving the monitoring system by playing pre-recorded videos, and enhances the system's ability to resist intrusion and tampering.

[0115] By detecting anomalies in video data, the system automatically activates a backup server for in-depth analysis when an anomaly is detected, through intelligent switching of the data switching switch. This ensures that monitoring and analysis can continue even in abnormal situations, guaranteeing that the monitoring system can continuously and accurately identify and warn of abnormal personnel. It also avoids monitoring blind spots caused by video replacement, ensuring the effective implementation of park security measures.

[0116] Example 3

[0117] like Figure 3 As shown in the figure, the intelligent campus security risk perception method with predictive capabilities in this embodiment includes:

[0118] Collect facial feature data from people registering at park access control;

[0119] Real-time acquisition of video data from T monitored areas within the park;

[0120] Extract real-time facial feature data from the video data appearing in the t-th monitoring area, t∈T. Compare the real-time facial feature data with the registered facial feature data to obtain the comparison results. The comparison results include whether there is an entry record or not. Mark the real-time facial feature data without an entry record as abnormal personnel features.

[0121] Pre-set designated areas to be avoided in each monitoring zone;

[0122] Using a tracking algorithm, we track abnormal individuals with unusual characteristics and record the areas in which these characteristics appear in q monitored areas. These areas are marked as "passing-through avoidance areas". We then connect all the recorded passing-through avoidance areas to form the movement trajectory of the abnormal individuals, where q is less than or equal to T.

[0123] Image frames of people with abnormal behavior are captured from video data, input into the action recognition model, and the avoidance actions of the abnormal people are obtained. The number of avoidance actions per unit time is counted.

[0124] Collect the current light intensity and humidity, and combine them with the number of avoidance actions to generate an avoidance action coefficient;

[0125] Obtain movement characteristic data and height data of abnormal individuals based on video data;

[0126] The movement trajectory, avoidance action coefficient, movement characteristic data, and height data are input into the behavioral characteristic analysis model to obtain the behavioral analysis results;

[0127] If the behavior analysis results indicate abnormal behavior, an alert will be issued.

[0128] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predictive intelligent park security risk perception, characterized in that, include: Collect facial feature data from people registering at park access control; Real-time acquisition of video data from T monitored areas within the park; Extract real-time facial feature data from the video data appearing in the t-th monitoring area, t∈T. Compare the real-time facial feature data with the registered facial feature data to obtain the comparison results. The comparison results include whether there is an entry record or not. Mark the real-time facial feature data without an entry record as abnormal personnel features. Pre-set designated areas to be avoided in each monitoring zone; Using a tracking algorithm, we track abnormal individuals with unusual characteristics and record the areas in which these characteristics appear in q monitored areas. These areas are marked as "passing-through avoidance areas". We then connect all the recorded passing-through avoidance areas to form the movement trajectory of the abnormal individuals, where q is less than or equal to T. Image frames of people with abnormal behavior are captured from video data, input into the action recognition model, and the avoidance actions of the abnormal people are obtained. The number of avoidance actions per unit time is counted. Collect the current light intensity and humidity, and combine them with the number of avoidance actions to generate an avoidance action coefficient; Obtain movement characteristic data and height data of abnormal individuals based on video data; The movement trajectory, avoidance action coefficient, movement characteristic data, and height data are input into the behavioral characteristic analysis model to obtain the behavioral analysis results; If the behavior analysis results indicate abnormal behavior, an alert will be issued; Also includes: Collect wind force and direction data for each monitored area; Based on wind force and direction data and light intensity, analyze whether the video data of each monitoring area is abnormal. If abnormal, generate an activation command; if normal, do not generate an activation command and send the activation command to the backup server and data switching switch. The backup server can perform real-time analysis of the video data. The backup server is activated according to the activation command, and the data switching switch transmits the video data to the backup server for analysis according to the activation command. Methods for analyzing whether video data in each monitored area is abnormal based on wind force and direction data and light intensity include: When the wind force level of the r-th monitoring area is greater than the preset wind force level, analyze the offset direction of the flexible ribbon in the video data of the r-th monitoring area to see if it is consistent with the wind direction. If it is inconsistent, generate an activation command; if it is consistent, do not generate an activation command. If the wind force level is less than or equal to the preset wind force level, the real-time collected light intensity and the real-time image frames from the video data of the r-th monitoring area will be input into the monitoring anomaly analysis model to determine whether the image frames are abnormal. If the image frames are abnormal, an activation command will be generated; if the image frames are normal, an activation command will not be generated. The training method for the monitoring anomaly analysis model includes: E sets of training data are collected in advance. Each set of training data includes the illumination intensity and the corresponding image frame, where E is an integer greater than 1. The training data is used as the input to the monitoring anomaly analysis model. The monitoring anomaly analysis model takes the illumination intensity corresponding to each set of image frames as the output and the actual illumination intensity corresponding to each set of image frames as the prediction target. The training objective is to minimize the sum of prediction errors for all illumination intensities. The monitoring anomaly analysis model is trained until the sum of prediction errors converges, at which point training stops. The monitoring anomaly analysis model uses a convolutional neural network.

2. The intelligent park security risk perception method with predictive capabilities according to claim 1, characterized in that, The avoidance actions include looking down and covering the face; the number of times the head is looked down and the face is covered is counted per unit time. The generation of the avoidance action coefficient includes: The avoidance action value is obtained by weighted summing of the number of times the head is lowered and the face is covered per unit time. The environmental impact value is obtained by weighted summing of the current light intensity and humidity. The avoidance action coefficient is obtained by dividing the avoidance action value by the environmental impact value.

3. The intelligent park security risk perception method with predictive capabilities according to claim 1, characterized in that, Methods for extracting movement characteristic data of abnormal individuals include: Motion characteristic data includes the average movement speed and the standard deviation of movement speed; The location coordinates of the abnormal personnel in continuous frames of video data are mapped to the electronic map of the park to obtain the continuous location coordinate sequence of the abnormal personnel within a unit time interval. Calculate the speed S moving per unit interval: X i+1 represents the abscissa of the (i+1)th position coordinate, X i represents the abscissa of the ith position coordinate, Y i+1 represents the ordinate of the (i+1)th position coordinate, Y i represents the ordinate of the ith position coordinate; The calculated movement speeds are used to create a set of movement speeds, and the average movement speed and standard deviation of the movement speeds within the set are calculated.

4. The intelligent park security risk perception method with predictive capabilities according to claim 1, characterized in that, Methods for obtaining height data of individuals with abnormal heights include: Using video processing techniques, the top and bottom pixel coordinates of reference objects and unusual persons are obtained in the video data image; The height of the reference object in the video data image is obtained based on the top and bottom pixel coordinates of the reference object. Similarly, the height of the abnormal person in the video data image can be obtained based on the top and bottom pixel coordinates of the abnormal person; The ratio of the height of the reference object in the video data image to the actual height of the reference object is marked as the scaling factor; the height of the abnormal person in the video data image is multiplied by the scaling factor to obtain the height data of the abnormal person.

5. The intelligent park security risk perception method with predictive capabilities according to claim 1, characterized in that, Training methods for behavioral feature analysis models include: K sets of action trajectories, avoidance action coefficients, height of abnormal personnel, and movement feature data are collected in advance and combined into K sets of feature vectors, where K is an integer greater than 1. Supervised learning methods are used, such as support vector machines, random forests, or deep neural networks, to train the behavior feature analysis model. Normal and abnormal behaviors are pre-labeled on the feature vectors to form a dataset of known behaviors. The dataset is divided into training and testing sets for training and validation of the behavior feature analysis model, and the output behavior feature analysis model that meets the preset accuracy is obtained.

6. The intelligent park security risk perception method with predictive capabilities according to claim 1, characterized in that, Methods for analyzing the offset direction of the flexible ribbon in the video data of the r-th monitoring area include: In the video data display, a pre-defined planar coordinate system is used, with the fixed end of the flexible ribbon as the origin of the planar coordinate system. The coordinates of each pixel in the area occupied by the flexible ribbon in the display at the current moment are obtained. The quadrant of the planar coordinate system where the flexible ribbon is located is determined based on the coordinates of each pixel. The offset direction of the flexible ribbon is determined based on the pre-defined direction of the quadrant.

7. A smart park security risk perception system with predictive capabilities, used to implement the smart park security risk perception method with predictive capabilities described in any one of 1-6, characterized in that, include: The first data acquisition module is used to collect facial feature data of people registering at the park's access control system and store it in the registration database. The second acquisition module is used to acquire video data from T monitored areas in the park in real time. The first analysis module extracts real-time facial feature data from the video data appearing in the t-th monitoring area, where t∈T. It compares the real-time facial feature data with the registered facial feature data to obtain the comparison results. The comparison results include whether there is an entry record or not. The real-time facial feature data without an entry record is marked as abnormal personnel features. The settings module allows you to pre-set the areas to be deliberately avoided in each monitoring zone; The second analysis module uses a tracking algorithm to track abnormal individuals with abnormal characteristics and records the areas deliberately avoided by these individuals in q monitoring areas. These areas are marked as "passing-through avoidance areas". All the recorded passing-through avoidance areas are connected sequentially to form the movement trajectory of the abnormal individuals, where q is less than or equal to T. The third analysis module extracts image frames of people with abnormal behavior from the video data, inputs them into the action recognition model, obtains the avoidance actions of the abnormal people, and counts the number of avoidance actions per unit time. The fourth analysis module collects the current light intensity and humidity, and combines this with the number of avoidance actions to generate an avoidance action coefficient. The fifth analysis module is used to obtain the movement characteristics and height data of abnormal persons based on video data; The comprehensive analysis module is used to input movement trajectory, avoidance action coefficient, movement characteristic data and height data into the behavior characteristic analysis model to obtain behavior analysis results; The early warning module issues an early warning if the behavior analysis results indicate abnormal behavior.

Images