Safety early warning method and device for abnormal behavior trajectory analysis and medium

Abstract

The invention provides a safety early warning method and device for abnormal behavior trajectory analysis and a medium, and relates to the technical field of safety early warning, and the method comprises the steps: collecting behavior trajectory data of a target region through a multi-source sensing device, and carrying out the multi-dimensional feature extraction of a behavior trajectory data set; according to the stay frequency characteristics, carrying out abnormity determination on the track space-time distribution characteristics and the speed change characteristics, and constructing an abnormal behavior determination vector; performing wandering mode recognition based on the abnormal behavior judgment vector, determining an abnormal behavior level, and triggering a graded early warning mechanism according to the abnormal behavior level; and generating an abnormal behavior alarm signal according to the grading early warning mechanism, combining the abnormal behavior alarm signal with the electronic fence information of the target area to generate a safety early warning report, and pushing the report to a monitoring terminal. According to the invention, the technical problem of abnormal behavior recognition accuracy in the prior art can be solved, and the technical effect of improving the behavior judgment accuracy is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of safety early warning, and in particular to a safety early warning method and device for abnormal behavior trajectory analysis and a medium. BACKGROUND

[0002] In the development process of current public safety governance, key area protection, and intelligent patrol systems, behavior trajectory analysis technology is gradually becoming a core support means for realizing proactive safety early warning and emergency response scheduling. Especially in the scenes of urban security monitoring, transportation hub safety management, enterprise and institution park prevention and control, etc., the behavior recognition and abnormal behavior detection system based on video monitoring and multi-source sensing equipment has been preliminarily applied. The existing technology usually extracts the walking path of the target personnel through the video stream, and then combines the target detection algorithm and the face recognition model to confirm the personnel identity, so as to assist the management personnel to find potential threat behaviors and improve the safety protection capability. However, the existing technology still has many limitations and key problems in actual application.

[0003] At present, the recognition of behavior trajectory of most current systems often relies on a single modal data source, such as video monitoring, infrared imaging, or access card recording, lacks multi-source sensing fusion means, and is difficult to obtain continuous, complete, and structured trajectory data. Such fragmented trajectory information is difficult to support high-precision judgment of complex behavior patterns and prediction of evolution trends. Secondly, most of the existing behavior discrimination methods are based on static rule matching or traditional machine learning models, which lack deep modeling capability for the mutual coupling relationship of time evolution, space transfer, and identity characteristics, resulting in problems such as lag, misjudgment, or omission in abnormal behavior detection. In addition, the current systems generally emphasize recognition and ignore reasoning, that is, even if an abnormal behavior such as loitering or abnormal residence is detected, there is still a lack of further causal reasoning and impact factor identification mechanism, which cannot support the priority response and resource scheduling in intelligent decision-making.

[0004] In summary, the existing technology has the technical problem that due to the reliance on a single modal data source, insufficient modeling capability, and lack of reasoning mechanism, the behavior trajectory information is fragmented, the accuracy and real-time performance of abnormal behavior recognition are low, and further, the efficient early warning and response scheduling capability of the system for potential risk behaviors in complex scenes is affected. SUMMARY

[0005] The purpose of the present application is to provide a safety early warning method and device for abnormal behavior trajectory analysis and a medium, to solve the technical problem in the existing technology that due to the reliance on a single modal data source, insufficient modeling capability, and lack of reasoning mechanism, the behavior trajectory information is fragmented, the accuracy and real-time performance of abnormal behavior recognition are low, and further, the efficient early warning and response scheduling capability of the system for potential risk behaviors in complex scenes is affected.

[0006] In view of the above problems, the present application provides a security warning method and device for abnormal behavior trajectory analysis and a medium.

[0007] In a first aspect, the present application provides a security warning method for abnormal behavior trajectory analysis, which is implemented by a security warning device for abnormal behavior trajectory analysis, and includes: collecting behavior trajectory data of a target area by a multi-source sensing device, performing multi-dimensional feature extraction on the behavior trajectory data set, wherein the multi-dimensional features include trajectory spatiotemporal distribution features, speed change features, and stay frequency features; performing abnormality determination on the trajectory spatiotemporal distribution features and the speed change features according to the stay frequency features, and constructing an abnormal behavior determination vector; performing wandering pattern recognition based on the abnormal behavior determination vector, determining an abnormal behavior level, triggering a hierarchical warning mechanism according to the abnormal behavior level; and generating an abnormal behavior warning signal according to the hierarchical warning mechanism, combining electronic fence information of the target area to generate a security warning report and pushing the report to a monitoring terminal.

[0008] Preferably, the security warning method for abnormal behavior trajectory analysis further includes: performing data collection by traversing the target area through the multi-source sensing device to obtain a multi-source sensing data set, wherein the multi-source sensing data set includes continuous video stream data, target positioning sequences, and identity verification information; performing human body recognition based on the continuous video stream data to extract skeleton key point data; performing motion analysis based on the target positioning sequences to obtain motion feature data; performing identity identification based on the identity verification information to generate identity tag data; and performing spatiotemporal data fusion on the skeleton key point data, the motion feature data, and the identity tag data to generate the behavior trajectory data set.

[0009] Preferably, the security warning method for abnormal behavior trajectory analysis further includes: dividing the target area into a plurality of sub-networks, mapping the skeleton key point data to the plurality of sub-networks to obtain point density distribution data; performing residence calculation based on the point density distribution data to obtain a regional residence index, constructing a spatiotemporal distribution matrix, mapping the regional residence index to the spatiotemporal distribution matrix for feature analysis to obtain trajectory spatiotemporal distribution features; performing adjacent displacement calculation based on the motion feature data to obtain an instantaneous speed sequence; performing smoothing processing according to the instantaneous speed sequence to construct a speed change rate histogram, traversing the speed change rate histogram to monitor the frequency of continuous changes, and obtaining speed change features; performing stay analysis on the behavior trajectory data set according to the regional residence index combined with the speed change features to determine a plurality of stay hotspot areas; and drawing a stay frequency heat map based on the plurality of stay hotspot areas combined with the identity tag data, performing feature coding according to the stay frequency heat map, and determining the stay frequency features.

[0010] Preferably, the security early warning method of the abnormal behavior trajectory analysis further comprises: performing time decay analysis according to the stay frequency feature to generate a stay frequency index, the stay frequency index comprising a dynamic stay weight; performing region sensitivity correction on the trajectory spatio-temporal distribution feature based on the dynamic stay weight to generate a spatio-temporal anomaly coefficient; performing path analysis according to the spatio-temporal anomaly coefficient to determine a path complexity; performing time sequence correlation on the stay frequency feature and the speed change feature to obtain a time sequence correlation coefficient, performing speed fluctuation analysis based on the time sequence correlation coefficient to obtain a speed fluctuation coefficient; performing abnormality determination on the speed fluctuation coefficient according to the stay frequency index combined with the path complexity to obtain an abnormality determination result, performing multi-dimensional behavior processing on the abnormality determination result to construct the abnormal behavior determination vector.

[0011] Preferably, the security early warning method of the abnormal behavior trajectory analysis further comprises: calling historical behavior trajectory record logs of a target area, traversing the historical behavior trajectory record logs to extract normal trajectory samples, and constructing a normal behavior trajectory sample database; performing benchmark analysis based on the normal behavior trajectory sample database to determine trajectory feature benchmark distribution data; performing deviation analysis on the abnormal behavior determination vector according to the trajectory feature benchmark distribution data to determine an abnormal deviation degree; setting a deviation critical threshold, and activating a wandering mode when the abnormal deviation degree is greater than the deviation critical threshold, performing trajectory abnormality influence analysis through the wandering mode to determine a dominant influence factor set; performing grade division according to the dominant influence factor set to determine the abnormal behavior grade.

[0012] Preferably, the security early warning method of the abnormal behavior trajectory analysis further comprises: performing regression calculation based on the trajectory feature benchmark distribution data to construct a threshold baseline, and setting a deviation critical threshold according to the threshold baseline; comparing the abnormal deviation degree with the deviation critical threshold, and activating the wandering mode when the abnormal deviation degree is greater than the deviation critical threshold; performing influence and cause reasoning on the abnormal behavior determination vector through the wandering mode to determine a plurality of influence factors; performing dominant analysis on the plurality of influence factors to determine the dominant influence factor set.

[0013] Preferably, the security early warning method of the abnormal behavior trajectory analysis further comprises: extracting an abnormal trajectory segment through the hierarchical early warning mechanism to perform spatio-temporal backtracking and generate a three-dimensional trajectory reproduction graph; generating a suspicious person portrait based on the three-dimensional trajectory reproduction graph combined with the identity label data, performing behavior correlation analysis according to the suspicious person portrait to generate a behavior correlation graph; introducing a historical case library of a target area, performing similarity matching based on the behavior correlation graph combined with the historical case library to generate a risk probability evaluation value, and adding the risk probability evaluation value to the alarm signal.

[0014] Preferably, the security warning method of the abnormal behavior trajectory analysis further comprises: performing perimeter protection analysis on the target area, constructing a defense level matrix, protecting the target area based on the defense level matrix, and constructing electronic fence information; performing defense feasibility analysis on the warning signal according to the electronic fence information to generate a defense feasibility result, the defense feasibility result including active defense parameters and passive defense parameters; when the defense feasibility result is the active defense parameters, a first security warning report is generated, an information push priority is set based on the risk probability evaluation value, and the security warning report is pushed to the monitoring terminal according to the information push priority; when the defense feasibility result is the passive defense parameters, a second security warning report is generated and pushed to the monitoring terminal and an emergency personnel control instruction is started.

[0015] In a second aspect, the present application also provides a security warning device for abnormal behavior trajectory analysis, which is used to execute the security warning method of abnormal behavior trajectory analysis as described in the first aspect, and comprises: a behavior trajectory data collection module, which is used to collect behavior trajectory data of a target area through multi-source sensing devices, and perform multi-dimensional feature extraction on the behavior trajectory data set, the multi-dimensional features including trajectory spatio-temporal distribution features, speed change features, and stay frequency features; an abnormality determination module, which is used to perform abnormality determination on the trajectory spatio-temporal distribution features and the speed change features according to the stay frequency features, and construct an abnormal behavior determination vector; a loitering pattern recognition module, which is used to perform loitering pattern recognition based on the abnormal behavior determination vector, determine an abnormal behavior level, and trigger a hierarchical warning mechanism according to the abnormal behavior level; and a report push module, which is used to generate a security warning report based on the abnormal behavior warning signal and the electronic fence information of the target area, and push the security warning report to a monitoring terminal.

[0016] In a third aspect, a computer readable storage medium is provided, and the computer readable storage medium stores a computer program, which, when executed, implements the steps of the security warning method of abnormal behavior trajectory analysis according to any one of the first aspect.

[0017] The technical solutions provided in the present application have at least the following technical effects or advantages: by achieving the technical target of behavior trajectory holographic modeling and abnormal behavior intelligent recognition based on multi-source sensing fusion, the behavior determination accuracy is improved, the timeliness of abnormal warning is enhanced, and dynamic security response scheduling is supported.

[0018] The above description is only a summary of the technical solutions of the present application. In order to more clearly understand the technical means of the present application, the specific embodiments of the present application can be implemented according to the content of the specification, and in order to make the above and other purposes, characteristics and advantages of the present application more obvious and easy to understand, the following specific embodiments of the present application are described. It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are only exemplary, and other drawings can be obtained by those skilled in the art without creative labor on the basis of the provided drawings.

[0020] Figure 1 The flowchart of the safety warning method of the abnormal behavior trajectory analysis of the present application; Figure 2 The structural diagram of the safety warning device of the abnormal behavior trajectory analysis of the present application.

[0021] Explanation of reference signs: behavior trajectory data acquisition module 1, abnormality determination module 2, wandering mode identification module 3, report pushing module 4. DETAILED DESCRIPTION

[0022] The present application provides a safety warning method, device and medium for abnormal behavior trajectory analysis, which solves the technical problem in the prior art that due to the dependence on a single modal data source, insufficient modeling capability and lack of reasoning mechanism, the behavior trajectory information is fragmented, the accuracy and real-time performance of abnormal behavior recognition are low, and further affect the efficient early warning and response scheduling capability of the system to potential risk behaviors in complex scenarios. The technical target of behavior trajectory holographic modeling and intelligent identification of abnormal behaviors based on multi-source perception fusion is achieved, and the technical effects of improving behavior determination accuracy, enhancing abnormal warning timeliness and supporting dynamic safety response scheduling are achieved.

[0023] The technical solutions in the present application will be described clearly and completely below with reference to the drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments of the present application, and it should be understood that the present application is not limited by the example embodiments described here. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor belong to the scope of protection of the present application. In addition, it should be noted that only parts related to the present application are shown in the drawings for convenience of description, not all.

[0024] Embodiment one, please refer to the attached Figure 1 The application provides a security warning method for abnormal behavior trajectory analysis, applied to a security warning device for abnormal behavior trajectory analysis, and specifically includes the following steps: S1: Collect behavior trajectory data of a target area through a multi-source sensing device, and perform multi-dimensional feature extraction on the behavior trajectory data set, wherein the multi-dimensional features include trajectory spatiotemporal distribution features, speed change features, and stay frequency features.

[0025] Further, the application further includes: S11: data collection through multi-source sensing devices in a target area to obtain a multi-source sensing data set, wherein the multi-source sensing data set includes continuous video stream data, target positioning sequences, and identity verification information; S12: human body recognition based on the continuous video stream data to extract skeleton key point data; S13: motion analysis based on the target positioning sequences to obtain motion feature data; S14: identity identification based on the identity verification information to generate identity label data; and S15: spatiotemporal data fusion of the skeleton key point data, the motion feature data, and the identity label data to generate the behavior trajectory data set.

[0026] Further, the application further includes: S16: division of the target area into a plurality of sub-networks, mapping of the skeleton key point data to the plurality of sub-networks to obtain point density distribution data; S17: stay calculation based on the point density distribution data to obtain a regional stay index, construction of a spatiotemporal distribution matrix, mapping of the regional stay index to the spatiotemporal distribution matrix for feature analysis to obtain trajectory spatiotemporal distribution features; S18: adjacent displacement calculation based on the motion feature data to obtain an instantaneous speed sequence; S19: smoothing processing according to the instantaneous speed sequence to construct a speed change rate histogram, frequency monitoring of continuous changes through traversal of the speed change rate histogram to obtain speed change features; S110: stay analysis of the behavior trajectory data set according to the regional stay index in combination with the speed change features to determine a plurality of stay hotspot regions; and S111: drawing of a stay frequency heat map based on the plurality of stay hotspot regions in combination with the identity label data, feature coding according to the stay frequency heat map, and determination of the stay frequency features.

[0027] Specifically, the target area is the area to be analyzed for abnormal behavior trajectory. The multi-source perception device is a plurality of sensing devices, such as surveillance cameras, radars, infrared thermal imagers, Bluetooth positioning devices, etc. Data collection is performed by traversing the target area with the multi-source perception device, and all targets in the target area are collected by the plurality of sensing devices to obtain a multi-source perception dataset. The multi-source perception dataset includes continuous video stream data, target positioning sequence, and identity verification information. The continuous video stream data is a sequence of dynamic images obtained by uninterrupted shooting by devices such as surveillance cameras; the target positioning sequence is provided by GPS, UWB or WiFi positioning systems, and records the spatial position information of each target at different times; the identity verification information may include card swiping data, face recognition results or mobile phone MAC addresses, etc., for identifying and confirming the identity of the target individual.

[0028] Next, human body recognition is performed based on the continuous video stream data to determine whether there is a human body in the picture through image processing and computer vision technology. Human body recognition includes convolutional neural networks, YOLO series models, etc. After recognition, the skeleton key point data is further extracted, i.e. the coordinate information of the main joint positions such as the head, shoulders, knees, ankles, etc. in the human skeletal structure, which can help more accurately determine the posture, action and behavior of the person. Motion analysis is performed using the target positioning sequence. Motion analysis is to obtain the speed, acceleration, path direction, etc. feature information by calculating the position change of the target between consecutive time points, i.e. motion feature data. Identity identification is performed on the collected identity verification information by comparing the records in the database to confirm the identity of the target individual. When the human body recognition result and the identity identification result deviate within a preset threshold, a first-level risk scan is triggered, a preliminary alarm signal is generated and the behavior trajectory monitoring state is entered.

[0029] Finally, the skeleton key point data, motion feature data and identity tag data are spatio-temporally fused. Spatio-temporal fusion refers to integrating data of different times and different sources into a unified format data structure, which contains both time dimension and spatial coordinates and identity attributes, and finally generates a complete behavior trajectory dataset that can depict the movement path, behavior posture and identity background of each individual within a certain period of time, laying a foundation for subsequent behavior analysis and abnormal early warning. At the same time of generating the behavior trajectory dataset, the abnormal confidence value is dynamically calculated according to the data fluctuation amplitude, and when the confidence value exceeds the grading threshold, the multi-level early warning module is triggered to start the corresponding protection mechanism in a graded response manner. Table 1 is a partial record table of the latest behavior trajectory data.

[0030] Table 1: Partial record table of the latest behavior trajectory data

[0031] The target area is divided into multiple sub-networks, which refers to dividing the entire monitoring range according to a fixed spatial grid or logical area, such as dividing a shopping mall into multiple independent store areas or passageways. Then the obtained skeleton key point data is mapped into the sub-network, that is, the joint coordinate positions of the human body at different times are projected into specific sub-areas. By counting the number of key points in each sub-area, point density distribution data is obtained, which is used to reflect the intensity of human activity in a certain area within a certain time.

[0032] After obtaining the point density distribution, the residence calculation is performed based on the density data to judge the time and frequency of the person staying in a certain area, and the area residence index is obtained to quantify the aggregation intensity and duration of individuals or groups in a certain area. Then, a space-time distribution matrix is constructed to organize the area residence index in the time dimension and the space dimension to better show the use of different areas by people at different time periods. Through statistical and feature extraction analysis of the space-time distribution matrix, the trajectory space-time distribution features can be finally obtained to describe the movement distribution and activity frequency of people in the entire area. According to the trajectory space-time distribution features and the historical security pattern library, when the deviation exceeds the warning threshold, an alarm is automatically generated and pushed to the monitoring terminal according to the regional protection priority.

[0033] At the same time, the obtained motion feature data is used to calculate the adjacent displacement, that is, the distance of position change between two consecutive time points is calculated using the difference method to obtain the instantaneous speed sequence, which reflects the moving speed of the person at each time. In order to avoid extreme values in the instantaneous speed sequence caused by jitter or errors, the instantaneous speed sequence is smoothed to remove abnormal fluctuations, and a speed change rate histogram is constructed to statistically analyze the distribution of speed change amplitude. The speed change rate histogram is accessed in turn and analyzed to obtain the speed change feature, which helps to determine whether the person in the target is quickly crossing the area or exists in the behavior of lingering or stopping. If the speed change feature presents a continuous low-speed lingering mode, active defense parameters such as sound and light warning or real-time broadcast intervention are immediately called to realize the active protection response at the behavior level.

[0034] Next, the area residence index and the speed change feature are combined to perform residence analysis. For example, only when a person stays in a certain area for a long time and moves slowly, it is determined as an effective stopping point, and multiple stopping hot area are further determined, which represent the main activity positions of the person, such as the lobby, the elevator port or the specific exhibition stand, etc. For the stopping hot area, combined with the electronic fence information, the protection feasibility analysis is performed to generate the corresponding defense feasibility result, which is used for subsequent safety warning decision.

[0035] After determining the stay hotspot area, identity tag data is introduced to count the access times and stay duration of different identity individuals in the stay hotspot area, and a stay frequency heat map is drawn. The heat map is an image that represents activity frequency in color intensity, and the color value of the stay frequency heat map is dynamically adjusted according to the product of the access times and the stay duration. Finally, feature coding is performed based on the stay frequency heat map, that is, the stay frequency of each area in the heat map is converted into a standardized vector feature, thereby determining the stay frequency feature and providing an important input for subsequent abnormal behavior recognition. According to the feature coding result and the defense feasibility result, a final security warning report is generated, including a risk probability evaluation value and a push priority, which is pushed to the monitoring terminal according to the information classification strategy, realizing a complete security protection closed loop from detection, evaluation to hierarchical warning.

[0036] S2: Abnormality determination is performed on the trajectory spatiotemporal distribution feature and the speed change feature according to the stay frequency feature, and an abnormal behavior determination vector is constructed.

[0037] Further, the present application also includes: S21: Time decay analysis is performed according to the stay frequency feature, a stay frequency index is generated, and the stay frequency index includes a dynamic stay weight; S22: The trajectory spatiotemporal distribution feature is regionally sensitive corrected based on the dynamic stay weight, and a spatiotemporal anomaly coefficient is generated; S23: Path analysis is performed according to the spatiotemporal anomaly coefficient, and path complexity is determined; S24: The stay frequency feature and the speed change feature are time-series associated to obtain a time-series association coefficient, speed fluctuation analysis is performed based on the time-series association coefficient to obtain a speed fluctuation coefficient; S25: Abnormality determination is performed on the speed fluctuation coefficient according to the stay frequency index combined with the path complexity, an abnormality determination result is obtained, and the abnormality determination result is processed in multiple dimensions to construct the abnormal behavior determination vector.

[0038] Specifically, time decay analysis is performed on the stay frequency feature. Time decay analysis refers to introducing a time dimension when counting the stay frequency of a person to a certain area, so that the weight of recent stay behavior is greater than that of past behavior, a stay frequency index is generated, the stay frequency index includes a dynamic stay weight, and reflects the strength of the individual's recent attention or stay to a certain area. The closer the time, the higher the weight. Based on the dynamic stay weight, multi-level warning determination can be automatically triggered. When the stay weight of an individual in a key area exceeds a preset hierarchical threshold, a first-level warning signal or a second-level warning signal is generated as a trigger basis for behavior anomaly monitoring.

[0039] After obtaining the dynamic stay weight, the dynamic stay weight is applied to the trajectory spatiotemporal distribution feature, and each region is corrected by weight, and the region appearing frequently in the recent period is given a higher sensitive coefficient to highlight the key region in the individual recent behavior. The spatiotemporal anomaly coefficient generated after correction is used to measure whether the person shows an abnormal activity pattern different from most people at a specific time and place. In combination with the electronic fence information and the regional protection level matrix, the region with a spatiotemporal anomaly coefficient exceeding the warning threshold is marked as a key protection area, and a monitoring terminal is automatically dispatched to perform alarm pushing.

[0040] After obtaining the spatiotemporal anomaly coefficient, path analysis is further carried out, which refers to identifying the moving route of an individual in a target region and judging whether there is an abnormal detour, frequent return or irregular crossing behavior. According to the complexity of the behavior, the path complexity is calculated to reflect the non-linearity and change degree of the path. When the path complexity reaches a preset threshold, a behavior anomaly grading response is triggered, and active defense or passive defense measures can be executed according to the grading setting.

[0041] Subsequently, the stay frequency feature and the speed change feature are time-series correlated, i.e., the two types of data are paired and analyzed on the time axis, and the time-series correlation coefficient is calculated to measure whether the speed change is synchronized with the stay behavior. For example, if a person slows down and frequently stays in a certain time period, it indicates that the two behaviors may be related. Through speed fluctuation analysis, a speed fluctuation coefficient is obtained, which represents the speed stability of a person during movement. The greater the fluctuation, the more abnormal the behavior. The speed fluctuation coefficient, in combination with the dynamic stay weight, can further determine the warning level, and the abnormal behavior is pushed to the monitoring terminal or triggers an automatic intervention device, such as an audible and visual warning or access control restriction, according to the set grading.

[0042] Finally, the stay frequency index, path complexity and speed fluctuation coefficient are jointly analyzed to determine the abnormal behavior, and the abnormal behavior determination result is obtained. The abnormal behavior determination result is further processed in multiple dimensions, i.e., more dimensional data (such as identity information, time characteristics, historical behavior, etc.) are fused to construct an abnormal behavior determination vector, which can be used as the final input for security warning or behavior portrait modeling. The abnormal behavior determination vector, in combination with the electronic fence constraint and the grading warning strategy, generates a final security warning report, including the risk level, warning grading and pushing priority, and can trigger active or passive defense measures to realize a complete closed loop from behavior anomaly detection to security response.

[0043] S3: Based on the abnormal behavior determination vector, a wandering mode is identified, an abnormal behavior level is determined, and a grading warning mechanism is triggered according to the abnormal behavior level.

[0044] Further, the application further comprises: S31: calling the historical behavior trajectory record log of the target area, traversing the historical behavior trajectory record log to extract normal trajectory samples, and constructing a normal behavior trajectory sample database; S32: performing benchmark analysis based on the normal behavior trajectory sample database to determine trajectory feature benchmark distribution data; S33: performing deviation analysis on the abnormal behavior judgment vector according to the trajectory feature benchmark distribution data to determine an abnormal deviation degree; S34: setting a deviation critical threshold, and when the abnormal deviation degree is greater than the deviation critical threshold, activating a wandering mode, performing trajectory abnormality influence analysis through the wandering mode, and determining a dominant influence factor set; and S35: performing hierarchical division according to the dominant influence factor set to determine the abnormal behavior level.

[0045] Further, the application further comprises: S341: performing regression calculation based on the trajectory feature benchmark distribution data to construct a threshold baseline, and setting a deviation critical threshold according to the threshold baseline; S342: comparing and judging the abnormal deviation degree according to the deviation critical threshold, and when the abnormal deviation degree is greater than the deviation critical threshold, activating the wandering mode; S343: performing influence and cause reasoning on the abnormal behavior judgment vector through the wandering mode to determine a plurality of influence factors; and S344: performing dominant analysis on the plurality of influence factors to determine the dominant influence factor set.

[0046] Specifically, the historical behavior trajectory record log of the target area is called. The historical behavior trajectory record log refers to the movement data of personnel in the target area within a historical time, including time stamp, position information, speed change, stay time and the like. By sequentially accessing the historical behavior trajectory record log, data samples conforming to normal behavior characteristics, i.e., normal trajectory samples, can be screened therefrom, which are used to construct a standard reference normal behavior trajectory sample database. The safety level information of each trajectory sample is labeled in the normal behavior trajectory sample database, so as to be referenced by the historical risk level when the abnormal behavior judgment triggers an early warning subsequently.

[0047] Statistical modeling is performed based on the normal behavior trajectory sample database to form trajectory feature benchmark distribution data. The trajectory feature benchmark distribution data refers to the probability distribution or numerical range established for multiple behavior characteristics such as path complexity, residence frequency, speed fluctuation, and the like, so as to obtain the typical behavior mode of normal people in the target area. Based on the trajectory feature benchmark distribution data, hierarchical threshold values can be set for corresponding abnormal behavior level early warning triggers, for example, three levels of low, medium and high.

[0048] Next, the obtained abnormal behavior judgment vector is compared with the trajectory feature reference distribution data to perform deviation analysis, and a deviation degree is calculated as an abnormal deviation degree. The abnormal deviation degree is a measure of the difference between the current behavior and the historical normal behavior. When the abnormal deviation degree exceeds a preset threshold, a hierarchical early warning mechanism is automatically triggered, the warning level is determined according to the deviation amplitude, and a preliminary alarm signal is generated.

[0049] Based on the trajectory feature reference distribution data, a regression calculation is performed to construct a dynamically changing threshold baseline. Regression calculation is a mathematical modeling process that can functionally fit the trajectory feature reference distribution data to construct a baseline for judging whether the behavior is normal, i.e., a threshold baseline. Through the threshold baseline, the threshold baseline can be set according to the standard behavior in different situations, and the deviation critical threshold is set to measure the boundary of the degree of behavior abnormality, for example, when the path complexity is greater than the set critical value, it is considered to deviate from the normal behavior. The threshold baseline combined with the electronic fence protection level matrix can dynamically adjust the deviation critical threshold of different regions, making the early warning more accurate.

[0050] Next, the deviation critical threshold is used to compare and judge the abnormal deviation degree of the current behavior. If the abnormal deviation degree exceeds the set deviation critical threshold, the wandering mode in the behavior monitoring mechanism can be triggered. The wandering mode refers to a working state that switches to more detailed analysis of abnormal behavior, which is used to identify risk behaviors such as non-purpose movement and continuous stay. Triggering the wandering mode generates corresponding hierarchical warning signals, and determines whether to push to the monitoring terminal or trigger safety intervention according to the preset active defense or passive defense strategy.

[0051] In the wandering mode, the factors involved in the current abnormal behavior are further analyzed, and the influence causal reasoning is performed on the abnormal behavior judgment vector. The influence causal reasoning includes the judgment factors of spatial factors (electronic fence distance, monitoring blind area coverage), time factors (time period abnormality coefficient, duration ratio) and behavior factors (speed variation degree, path fractal dimension) to determine multiple influence factors. According to the identification results of the influence factors, the electronic fence response level, the monitoring area rotation priority or the behavior intervention strategy is adjusted to realize closed-loop early warning.

[0052] Finally, a dominant analysis is performed on all identified impact factors, i.e., determining the factors that have the greatest impact on the current abnormal behavior. Dominant analysis usually involves statistical weight evaluation, information gain ranking or feature contribution rate analysis, and finally extracts a dominant impact factor set from multiple impact factors. For example, when the space factor is dominant, the electronic fence defense level is raised and the monitoring angle layout is optimized, when the time factor is dominant, the patrol frequency and response preparation level of the period are adjusted, and when the behavior factor is dominant, the deep learning model is started for cross-scene abnormal pattern retrieval. The dominant impact factor set is also used to determine the warning strategy type, such as active defense trigger, passive defense record or mixed mode execution.

[0053] Finally, the dominant impact factor set is used for level division. Level division refers to dividing the abnormal degree of behavior into several levels, such as low risk, medium risk and high risk, forming a final abnormal behavior level, which is then used in the dynamic warning system to trigger the corresponding response mechanism.

[0054] The hierarchical warning mechanism is triggered according to the abnormal behavior level. The hierarchical warning mechanism refers to a system that automatically takes corresponding response measures for different abnormal behavior levels, including information prompts, warning reports, behavior locking, linkage control, etc. Each level in the abnormal behavior level represents the risk degree or event urgency degree that the behavior may bring, for example, an abnormal residence time is longer but the behavior is relatively stable and can be classified as medium risk, while frequent speed changes, path rotation abnormalities and proximity to sensitive areas may be marked as high risk.

[0055] S4: generating an abnormal behavior warning signal according to the hierarchical warning mechanism, combining the electronic fence information of the target area to generate a security warning report and pushing it to the monitoring terminal.

[0056] Further, the present application also includes: S41: extracting an abnormal trajectory segment through the hierarchical warning mechanism for time and space backtracking, generating a three-dimensional trajectory reconstruction graph; S42: generating a suspicious person portrait based on the three-dimensional trajectory reconstruction graph combined with the identity tag data, performing behavior correlation analysis according to the suspicious person portrait, and generating a behavior correlation graph; S43: introducing a historical case library of the target area, performing similarity matching based on the behavior correlation graph combined with the historical case library, generating a risk probability evaluation value, and adding the risk probability evaluation value to the warning signal.

[0057] Further, the application further comprises: S44: performing perimeter protection analysis on the target area, constructing a defense level matrix, and performing protection on the target area based on the defense level matrix to construct electronic fence information; S45: performing defense feasibility analysis on the alarm signal according to the electronic fence information to generate a defense feasibility result, wherein the defense feasibility result comprises active defense parameters and passive defense parameters; S46: when the defense feasibility result is the active defense parameters, a first safety warning report is generated, an information push priority is set based on the risk probability evaluation value, and the safety warning report is pushed to a monitoring terminal according to the information push priority; and S47: when the defense feasibility result is the passive defense parameters, a second safety warning report is generated and pushed to the monitoring terminal and an emergency personnel control instruction is started.

[0058] Specifically, when the abnormal behavior level is detected and the hierarchical warning mechanism is triggered, spatiotemporal backtracking analysis is performed on the relevant abnormal trajectory segment. The abnormal trajectory segment refers to the moving path of a person who is determined to be abnormal within a certain period of time. Spatiotemporal backtracking refers to the mapping of the abnormal trajectory segment back to the original spatiotemporal coordinate system, so as to analyze the behavior development process from the two dimensions of time and space, and generate a three-dimensional trajectory reconstruction graph, i.e., the moving trajectory of the individual at a specific location and time is restored in a three-dimensional model, making the behavior path more intuitive.

[0059] Then, the three-dimensional trajectory reconstruction graph is combined with the identity label data to construct a portrait, and a suspicious person portrait can be generated, i.e., the features presented by the individual in the suspicious behavior are comprehensively depicted. Thereafter, behavior correlation analysis is carried out according to the suspicious person portrait, i.e., the relationship between the individual and other behaviors, persons or events is found out, and a behavior correlation graph is constructed. The behavior correlation graph is a multi-node and multi-edge structure, which describes the mutual influence between individual behaviors, for example, there is a trajectory overlap between a person and another person before entering a specific area, or other individuals with similar paths appear before and after a certain behavior occurs.

[0060] Subsequently, a historical case library of the target area is introduced, and the historical case library is a database storing past known behavior events and related trajectories, graphs, and conclusions. Similarity matching is performed between the current behavior correlation graph and the historical case library, i.e., the similarity between the current situation and the past events in the trajectory form, behavior mode, and character features is judged. The result calculated is a risk probability evaluation value, which is used to quantify whether the current behavior is likely to trigger a security event. Finally, the risk probability evaluation value is attached to the alarm signal, so that the alarm information no longer only contains whether it is abnormal, but also contains a quantitative judgment of the risk severity.

[0061] Before the target area is protected, a perimeter protection analysis is carried out, and the perimeter protection analysis refers to identification and evaluation of risk paths that the boundary of the target area may be invaded, approached or damaged. For example, the boundary of a construction site may include positions such as a fence, an entrance, a visual blind area and the like. Subsequently, a defense level matrix is constructed, including perimeter defense, three-level protection parameters of stereoscopic monitoring and biological blocking, for representing the level division of different regions or passages in security protection, which can be divided according to the risk level, the passage frequency or the historical alarm condition. Then, the entire target area is protected based on the defense level matrix, and finally, electronic fence information is generated, that is, logical defense lines are set on the boundary of the target area by virtual means, and the invasion or crossing behavior inside and outside the electronic fence is monitored by using a sensing device.

[0062] Then, according to the electronic fence information, defense feasibility analysis is carried out on the triggered alarm signal to determine whether a corresponding defense measure can be successfully activated under the current alarm condition. The output of the analysis is a defense feasibility result, which includes two types of active defense parameters and passive defense parameters.

[0063] The active defense parameter refers to a protective measure that can be actively taken, and the active defense proves that the current anomaly is within a controllable range, knows the abnormal trajectory, knows the identity of the abnormal personnel and the like. The passive defense parameter indicates that the current anomaly is outside the controllable range and waits for manual intervention. When it is judged that the defense feasibility result is the active defense parameter, a first security warning report is generated, which summarizes details of the alarm event, a risk level, associated trajectory information and the like, and sets an information push priority according to a risk probability evaluation value. The information push priority represents the alarm speed and transmission mode of the first security warning report. Conversely, when it is judged that the defense feasibility result is the passive defense parameter, a second security warning report is generated, and it is immediately pushed to a monitoring terminal. At the same time, an emergency personnel control instruction is started to dispatch on-site security personnel to the incident area, set an enclosed path, or notify the superior department to implement the next step of decision-making.

[0064] In summary, the security warning method for abnormal behavior trajectory analysis provided in the present application has the following technical effects: by achieving the technical targets of behavior trajectory holographic modeling and abnormal behavior intelligent identification based on multi-source sensing fusion, the technical effects of improving behavior determination accuracy, enhancing abnormal warning timeliness and supporting dynamic security response scheduling are achieved.

[0065] Embodiment two, based on the same inventive concept as the security warning method for abnormal behavior trajectory analysis in the foregoing embodiments, the present application also provides a security warning device for abnormal behavior trajectory analysis, please refer to the accompanying Figure 2The device comprises a behavior trajectory data collection module 1, which is used to collect behavior trajectory data of a target area through a multi-source perception device, and to perform multi-dimensional feature extraction on the behavior trajectory data set, wherein the multi-dimensional features include trajectory spatiotemporal distribution features, speed change features, and stay frequency features; an anomaly determination module 2, which is used to perform anomaly determination on the trajectory spatiotemporal distribution features and the speed change features according to the stay frequency features, and to construct an abnormal behavior determination vector; a loitering pattern recognition module 3, which is used to perform loitering pattern recognition based on the abnormal behavior determination vector, to determine an abnormal behavior level, and to trigger a hierarchical early warning mechanism according to the abnormal behavior level; and a report pushing module 4, which is used to generate an abnormal behavior warning signal according to the hierarchical early warning mechanism, to generate a security warning report in combination with electronic fence information of the target area, and to push the report to a monitoring terminal.

[0066] Further, the security warning device for abnormal behavior trajectory analysis is also used to: traverse a target area through a multi-source perception device to collect data, and obtain a multi-source perception data set, wherein the multi-source perception data set includes continuous video stream data, target positioning sequences, and identity verification information; perform human body recognition based on the continuous video stream data, and extract skeleton key point data; perform motion analysis based on the target positioning sequences, and obtain motion feature data; perform identity identification based on the identity verification information, and generate identity tag data; and perform spatiotemporal data fusion on the skeleton key point data, the motion feature data, and the identity tag data, and generate the behavior trajectory data set.

[0067] Further, the security warning device for abnormal behavior trajectory analysis is also used to: divide the target area into a plurality of sub-networks, map the skeleton key point data to the plurality of sub-networks, and obtain point density distribution data; perform residence calculation based on the point density distribution data, obtain a regional residence index, construct a spatiotemporal distribution matrix, map the regional residence index to the spatiotemporal distribution matrix for feature analysis, and obtain trajectory spatiotemporal distribution features; perform adjacent displacement calculation based on the motion feature data, and obtain an instantaneous speed sequence; perform smoothing processing according to the instantaneous speed sequence, construct a speed change rate histogram, traverse the speed change rate histogram to monitor the frequency of continuous changes, and obtain speed change features; perform stay analysis on the behavior trajectory data set according to the regional residence index in combination with the speed change features, and determine a plurality of stay hotspot areas; draw a stay frequency heat map based on the plurality of stay hotspot areas in combination with the identity tag data, perform feature coding according to the stay frequency heat map, and determine the stay frequency features.

[0068] Further, the security warning device for abnormal behavior trajectory analysis is further configured to: perform time decay analysis according to the stay frequency feature to generate a stay frequency index, wherein the stay frequency index comprises a dynamic stay weight; perform region sensitivity correction on the trajectory spatiotemporal distribution feature based on the dynamic stay weight to generate a spatiotemporal anomaly coefficient; perform path analysis according to the spatiotemporal anomaly coefficient to determine a path complexity; perform time sequence correlation on the stay frequency feature and the speed change feature to obtain a time sequence correlation coefficient; perform speed fluctuation analysis based on the time sequence correlation coefficient to obtain a speed fluctuation coefficient; perform abnormality determination on the speed fluctuation coefficient according to the stay frequency index and the path complexity to obtain an abnormality determination result; and perform multi-dimensional behavior processing on the abnormality determination result to construct the abnormal behavior determination vector.

[0069] Further, the security warning device for abnormal behavior trajectory analysis is further configured to: call historical behavior trajectory record logs of a target area, traverse the historical behavior trajectory record logs to extract normal trajectory samples, and construct a normal behavior trajectory sample database; perform benchmark analysis based on the normal behavior trajectory sample database to determine trajectory feature benchmark distribution data; perform deviation analysis on the abnormal behavior determination vector according to the trajectory feature benchmark distribution data to determine an abnormal deviation degree; set a deviation critical threshold value, and activate a loitering mode when the abnormal deviation degree is greater than the deviation critical threshold value; perform trajectory abnormality influence analysis through the loitering mode to determine a dominant influence factor set; and perform grade division according to the dominant influence factor set to determine the abnormal behavior grade.

[0070] Further, the security warning device for abnormal behavior trajectory analysis is further configured to: perform regression calculation based on the trajectory feature benchmark distribution data to construct a threshold baseline, set a deviation critical threshold value according to the threshold baseline; compare and determine the abnormal deviation degree according to the deviation critical threshold value, and activate the loitering mode when the abnormal deviation degree is greater than the deviation critical threshold value; perform influence and cause reasoning on the abnormal behavior determination vector through the loitering mode to determine a plurality of influence factors; and perform dominant analysis on the plurality of influence factors to determine the dominant influence factor set.

[0071] Further, the security warning device for abnormal behavior trajectory analysis is further configured to: perform spatiotemporal backtracking on an abnormal trajectory segment extracted through the hierarchical warning mechanism to generate a three-dimensional trajectory reproduction graph; perform portrait construction based on the three-dimensional trajectory reproduction graph and the identity label data to generate a suspicious person portrait, perform behavior correlation analysis according to the suspicious person portrait to generate a behavior correlation graph, introduce a historical case library of a target area, perform similarity matching based on the behavior correlation graph and the historical case library to generate a risk probability evaluation value, and add the risk probability evaluation value to the warning signal.

[0072] Further, the security warning device for analyzing abnormal behavior trajectory is further configured to: perform perimeter protection analysis on the target area, construct a defense level matrix, perform protection on the target area based on the defense level matrix, and construct electronic fence information; perform defense feasibility analysis on the warning signal according to the electronic fence information, and generate a defense feasibility result, wherein the defense feasibility result includes active defense parameters and passive defense parameters; when the defense feasibility result is the active defense parameters, a first security warning report is generated, an information push priority is set based on the risk probability evaluation value, and the security warning report is pushed to a monitoring terminal according to the information push priority; and when the defense feasibility result is the passive defense parameters, a second security warning report is generated and pushed to the monitoring terminal and an emergency personnel control instruction is started.

[0073] The embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The security warning device for analyzing abnormal behavior trajectory in the foregoing embodiment one is also applicable to the security warning device for analyzing abnormal behavior trajectory in the present embodiment. The security warning device for analyzing abnormal behavior trajectory in the present embodiment can be clearly understood by the person skilled in the art based on the foregoing detailed description of the security warning method for analyzing abnormal behavior trajectory. Therefore, for the sake of brevity of the specification, the security warning device for analyzing abnormal behavior trajectory in the present embodiment will not be described in detail.

[0074] Embodiment three, based on the security warning method for analyzing abnormal behavior trajectory in the foregoing embodiments, the present application also provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program implements the steps of the security warning method for analyzing abnormal behavior trajectory in any one of the foregoing embodiments when executed.

[0075] The above description of the disclosed embodiments enables a person skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0076] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application also intends to include these modifications and variations.

Claims

1. A security early warning method for abnormal behavior trajectory analysis, characterized in that, The method includes: Behavioral trajectory data of the target area is collected by multi-source sensing devices, and multi-dimensional features are extracted from the behavioral trajectory dataset. The multi-dimensional features include trajectory spatiotemporal distribution features, speed change features, and dwell frequency features. Based on the dwell frequency characteristics, anomaly detection is performed on the trajectory spatiotemporal distribution characteristics and the velocity change characteristics to construct an abnormal behavior detection vector, including: Based on the dwell frequency characteristics, a time decay analysis is performed to generate a dwell frequency index, which includes a dynamic dwell weight. Based on the dynamic dwell weight, the spatiotemporal distribution characteristics of the trajectory are corrected for regional sensitivity, and a spatiotemporal anomaly coefficient is generated. Path analysis is performed based on the spatiotemporal anomaly coefficients to determine the path complexity; The dwell frequency feature and the speed change feature are correlated in time to obtain the time correlation coefficient. Based on the time correlation coefficient, speed fluctuation analysis is performed to obtain the speed fluctuation coefficient. The speed fluctuation coefficient is anomaly determined by combining the dwell frequency index with the path complexity, and the anomaly determination result is obtained. The anomaly determination result is then subjected to multi-dimensional behavior processing to construct the anomaly behavior determination vector. Based on the abnormal behavior determination vector, a loitering pattern is identified to determine the level of abnormal behavior. A tiered early warning mechanism is then triggered according to the abnormal behavior level, including: Retrieve historical behavior trajectory records in the target area, iterate through the historical behavior trajectory records to extract normal trajectory samples, and construct a normal behavior trajectory sample database. Based on the normal behavior trajectory sample database, benchmark analysis is performed to determine the baseline distribution data of trajectory features; The deviation analysis is performed on the abnormal behavior judgment vector based on the trajectory feature baseline distribution data to determine the degree of abnormal deviation. A deviation threshold is set. When the abnormal deviation exceeds the deviation threshold, a wandering mode is activated. The trajectory anomaly impact analysis is performed through the wandering mode to determine the set of dominant influencing factors. The abnormal behavior level is determined by classifying the levels according to the set of dominant influencing factors. Based on the aforementioned hierarchical early warning mechanism, an abnormal behavior alarm signal is generated. Combined with the electronic fence information of the target area, a security early warning report is generated and pushed to the monitoring terminal.

2. The security early warning method for abnormal behavior trajectory analysis as described in claim 1, characterized in that, Methods for collecting behavioral trajectory data of a target area using multi-source sensing devices include: Data is collected by traversing the target area using multi-source sensing devices to obtain a multi-source sensing dataset, which includes continuous video stream data, target positioning sequence, and identity verification information. Human body recognition is performed based on the continuous video stream data, and skeleton key point data is extracted; Motion analysis is performed based on the target localization sequence to obtain motion feature data; Based on the authentication information, identity is identified, and identity tag data is generated; The skeleton key point data, the motion feature data, and the identity tag data are fused in a spatiotemporal manner to generate the behavior trajectory dataset.

3. The security early warning method for abnormal behavior trajectory analysis as described in claim 2, characterized in that, Multi-dimensional feature extraction is performed on the behavioral trajectory dataset, including trajectory spatiotemporal distribution features, speed change features, and dwell frequency features. The method includes: The target region is divided into multiple sub-networks, and the skeleton key point data is mapped to the multiple sub-networks to obtain point density distribution data. Based on the point density distribution data, the dwell time is calculated to obtain the regional dwell time index, a spatiotemporal distribution matrix is ​​constructed, and the regional dwell time index is mapped to the spatiotemporal distribution matrix for feature analysis to obtain the trajectory spatiotemporal distribution features. Based on the motion feature data, adjacent displacements are calculated to obtain an instantaneous velocity sequence; The instantaneous velocity sequence is smoothed to construct a velocity change rate histogram. The frequency of continuous changes is monitored by traversing the velocity change rate histogram to obtain velocity change characteristics. Based on the regional dwelling index and the speed change characteristics, a dwelling analysis was performed on the behavioral trajectory dataset to identify multiple dwelling hotspot areas; A heatmap of dwell frequency is drawn based on the multiple dwelling hotspot areas and the identity tag data. Feature encoding is performed based on the dwell frequency heatmap to determine the dwell frequency feature.

4. The security early warning method for abnormal behavior trajectory analysis as described in claim 1, characterized in that, A deviation threshold is set. When the abnormal deviation exceeds the deviation threshold, a wandering mode is activated. Trajectory anomaly impact analysis is performed using this wandering mode to determine the set of dominant influencing factors. The method includes: Regression calculations are performed based on the trajectory feature baseline distribution data to construct a threshold baseline, and a deviation critical threshold is set according to the threshold baseline. The abnormal deviation is compared and judged according to the deviation threshold. When the abnormal deviation is greater than the deviation threshold, the lingering mode is activated. By using the lingering pattern to perform causal reasoning on the abnormal behavior judgment vector, multiple influencing factors are identified. The dominant influencing factors are analyzed to determine the set of dominant influencing factors.

5. A security early warning method for abnormal behavior trajectory analysis as described in claim 2, characterized in that, The method for generating abnormal behavior alarm signals based on the aforementioned hierarchical early warning mechanism includes: The abnormal trajectory segments are extracted through the hierarchical early warning mechanism for spatiotemporal backtracking, generating a three-dimensional trajectory reconstruction map. Based on the three-dimensional trajectory reconstruction map and the identity tag data, a profile is constructed to generate a profile of a suspicious person. Based on the profile of the suspicious person, a behavioral association analysis is performed to generate a behavioral association map. A historical case library of the target area is introduced, and similarity matching is performed based on the behavior association map and the historical case library to generate a risk probability assessment value. The risk probability assessment value is then added to the alarm signal.

6. The security early warning method for abnormal behavior trajectory analysis as described in claim 5, characterized in that, Abnormal behavior alarm signals are combined with electronic fence information of the target area to generate a security warning report and push it to the monitoring terminal. Methods include: Perform perimeter protection analysis on the target area, construct a defense level matrix, protect the target area based on the defense level matrix, and construct electronic fence information; Based on the electronic fence information, a defense feasibility analysis is performed on the alarm signal to generate a defense feasibility result, which includes active defense parameters and passive defense parameters. When the feasible defense result is the active defense parameter, a first security warning report is generated, and the information push priority is set based on the risk probability assessment value. The security warning report is then pushed to the monitoring terminal according to the information push priority. If the feasible result of the defense is the passive defense parameter, a second security warning report is generated, pushed to the monitoring terminal, and an emergency personnel control command is initiated.

7. A security early warning device for analyzing abnormal behavior trajectories, characterized in that, The steps for implementing the security early warning method for abnormal behavior trajectory analysis according to any one of claims 1 to 6 include: The behavior trajectory data acquisition module is used to collect behavior trajectory data of the target area through multi-source sensing devices, and to extract multi-dimensional features from the behavior trajectory dataset. The multi-dimensional features include trajectory spatiotemporal distribution features, speed change features, and dwell frequency features. Anomaly detection module is used to detect anomalies in the trajectory spatiotemporal distribution features and the speed change features based on the dwell frequency features, and to construct an abnormal behavior detection vector. The loitering pattern recognition module is used to recognize loitering patterns based on the abnormal behavior judgment vector, determine the level of abnormal behavior, and trigger a graded early warning mechanism according to the level of abnormal behavior. The report push module is used to generate an abnormal behavior alarm signal based on the hierarchical early warning mechanism, combine it with the electronic fence information of the target area to generate a safety early warning report, and push it to the monitoring terminal.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed, implements the steps of a security early warning method for abnormal behavior trajectory analysis as described in any one of claims 1 to 6.

Images