A security area intrusion warning method and system based on video images

Abstract

The present invention discloses a security area intrusion warning method and system based on video images, which belongs to the field of intrusion warning technology; through monitoring cameras distributed in the security area, video image data is collected in real time, the trajectory of the intrusion target is recorded using a preset target detection algorithm, the movement direction, speed change and residence time of the intrusion target are analyzed, and a dynamic motion trajectory is generated. In combination with the scene layout in the area, the intrusion path is inferred using a trajectory prediction model; the danger level of different potential paths is evaluated, and the paths are sorted by priority; according to the path priority, the warning sensitivity of the monitoring area is adjusted in real time, and if the path analysis result is a high-risk area, the warning system is triggered and an alarm message is issued. The present invention effectively identifies high-risk paths through comprehensive risk assessment and dynamic path sorting; and based on the sensitivity dynamic adjustment strategy and multi-level warning mechanism, the real-time and intelligence of the monitoring system are significantly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intrusion warning, and in particular to a method and system for early warning of intrusion in a security area based on video images. Background Art

[0002] With the rapid development of modern security technology, regional monitoring based on video images has been widely used in various scenarios, including traffic management, public place safety, industrial parks and residential communities. In these fields, real-time monitoring by installing high-resolution cameras combined with target detection and behavior analysis technology has become an important means to improve security efficiency. In recent years, with the advancement of deep learning and target recognition algorithms, the accuracy and speed of dynamic target detection have been significantly improved, and technologies such as trajectory analysis and path prediction have become increasingly mature, which provides a technical basis for achieving more accurate regional intrusion warnings. However, traditional monitoring systems mainly rely on simple motion detection and static rule settings, lacking the ability to intelligently analyze dynamic changes in complex scenarios, especially in path prediction and risk assessment. In addition, in scenarios with high real-time requirements, traditional systems find it difficult to balance the accuracy and computational efficiency of target detection, resulting in the inability to timely warn of potential threats.

[0003] Existing technologies still face many deficiencies in application: First, traditional systems cannot effectively cope with multi-target dynamic environments. Their performance in complex scenarios relies on manually set rules and lacks intelligent analysis of target trajectories and regional characteristics, making it difficult to generate accurate risk assessment results. Secondly, path prediction methods are mostly limited to simple straight line or fixed route models, ignoring the impact of dynamic characteristics such as target movement speed and acceleration as well as scene characteristics in the region, resulting in limited practicality of predicted paths. Finally, the warning mechanism is usually triggered based on a fixed threshold, and the sensitivity adjustment is not flexible enough, making it difficult to dynamically optimize the monitoring strategy based on real-time path risk. These technical bottlenecks make it difficult for existing security monitoring systems to provide efficient and accurate intrusion warning services in complex dynamic scenarios. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for early warning of intrusion in a security area based on video images, so as to solve the problems raised in the above-mentioned background technology.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] A security area intrusion warning method based on video images, the method comprises the following steps: step S1: through the surveillance cameras distributed in the security area, real-time video image data is collected, and basic formatting processing is performed on the data; step S2: using the preset target detection algorithm, the intrusion target is extracted from the video image, the trajectory of the intrusion target is recorded, the movement direction, speed change and residence time of the intrusion target are analyzed, and the dynamic movement trajectory is generated; step S3: based on the dynamic movement trajectory, combined with the scene layout in the area, the intrusion path is inferred using the trajectory prediction model; the danger level of different potential paths is evaluated, focusing on the paths close to the core area or with abnormal movement directions, and the paths are sorted according to priority; step S4: according to the path priority, the warning sensitivity of the monitoring area is adjusted in real time, such as increasing the detection frequency for high-risk paths or enhancing the detection threshold of the algorithm. When the path analysis result shows that the target may be close to the high-risk area, the early warning system is triggered, an alarm message is issued, and the path data is pushed to the security personnel at the same time.

[0007] As a preferred solution of the video image-based security area intrusion warning method of the present invention, the specific implementation process of step S2 includes:

[0008] Step S2.1: Based on the basic formatting processing result of step S1, a hierarchical target detection algorithm is used to extract dynamic targets from the video image, distinguish between background static elements and target dynamic features, and generate initial detection data of the dynamic targets, as follows:

[0009]

[0010] in, represents the initial detection data of target i, represents the weight of the nth feature, represents the data value of target i on the nth feature, Indicates the number of target features.

[0011] Step S2.2: Based on the initial detection data , using the spatial correlation of multiple cameras in the area, matching the position changes of dynamic targets in different perspectives, and constructing a multi-perspective spatial motion model of the dynamic target, as follows:

[0012]

[0013] in, is the multi-view spatial motion model of the dynamic target k, is the initial detection data of target k calculated in the cth camera, is the mean of the initial detection data of target k in all cameras, and C is the total number of cameras.

[0014] As a preferred solution of the video image-based security area intrusion warning method of the present invention, the specific implementation process of step S2 also includes:

[0015] Step S2.3: Based on the multi-view spatial motion model of the dynamic target, the real-time position of the dynamic target is recorded, and the speed change, acceleration and dwell time of the dynamic target are calculated in combination with the time series information to generate the trajectory data of the dynamic target, as follows:

[0016]

[0017]

[0018] in, is the velocity of target k at time t, is the acceleration of target k at time t, is the position of target k at time t, is the time interval, For target k at time location, For target k at time speed.

[0019] Step S2.4: Based on the trajectory data, match it with the scene feature database of the security area and correct the abnormal points in the trajectory, as follows:

[0020]

[0021] in, is the corrected trajectory point of target k at time t, is the multi-view spatial motion model of the dynamic target k, is the position of target k at time t, and Represents the multi-view spatial motion model of dynamic target k and the position of target k at time t The weight coefficient of .

[0022] Based on the corrected trajectory data, a dynamic motion trajectory dataset is generated, and the trajectory dataset is passed as input to step S3.

[0023] As a preferred solution of the video image-based security area intrusion warning method of the present invention, the specific implementation process of step S3 includes:

[0024] Step S3.1: Based on the dynamic motion trajectory data and the scene layout data of the security area, the cumulative risk value of the target path passing through different areas is calculated to obtain the preliminary path risk distribution data, as follows:

[0025]

[0026] in, is the initial risk value of the path segment m→n, is the scene feature value of the path segment m→n on feature p, is the length of the path segment m→n, is the weight of feature p, is the normalized coefficient of feature p, is the number of scene features.

[0027] Step S3.2: The preliminary path risk distribution data is integrated with the key monitoring characteristics of the region, the risk weight of the potential path is corrected, and the comprehensive risk value of the path segment m→n is generated as follows:

[0028]

[0029] in, is the comprehensive risk value of the corrected path segment m→n, is the weight coefficient of the key monitoring area, is the key monitoring feature value of the path segment m→n.

[0030] As a preferred solution of the video image-based security area intrusion warning method of the present invention, the specific implementation process of step S3 also includes:

[0031] Step S3.3: Using the path prediction model, the most likely intrusion path is identified from the regional risk matrix. At the same time, the path prediction priority is dynamically adjusted in combination with the real-time direction and speed changes of the target movement, as follows:

[0032]

[0033] in, is the priority of the path segment m→n, is the speed of the target at the starting node m of the path segment, is the straight-line distance of path segment m→n.

[0034] Step S3.4: Prioritize all potential paths, focus on paths close to high-risk areas or with abnormal target behaviors, and output path ranking data.

[0035] As a preferred solution of the video image-based security area intrusion warning method of the present invention, the specific implementation process of step S4 includes:

[0036] Step S4.1: According to the priority data of the predicted path, the monitoring sensitivity of the security area is dynamically adjusted as follows:

[0037]

[0038] in, is the monitoring sensitivity after dynamic adjustment, As the basic monitoring sensitivity, is the sensitivity adjustment factor, is the priority of the path segment m→n, The maximum value of all path priorities.

[0039] Step S4.2: For high-priority path segments, the accuracy of the image processing algorithm is increased, and the path priority is updated in real time by further analyzing the subtle features of the target (such as posture changes and speed fluctuations), as follows:

[0040]

[0041] in, is the update weight of the target feature, is the real-time speed fluctuation value of the target, is the target attitude change rate, is the speed fluctuation weight coefficient, is the weight coefficient of posture change.

[0042] As a preferred solution of the video image-based security area intrusion warning method of the present invention, the specific implementation process of step S4 also includes:

[0043] Step S4.3: When the path priority exceeds the preset threshold, the multi-level warning mechanism is triggered to output the alarm level and related target data.

[0044] Step S4.4: Push the alarm information and path data to the security personnel in real time, and generate a visual risk map showing the target's real-time location, movement trajectory and predicted path.

[0045] A security area intrusion warning system based on video images, the system comprises: a monitoring video acquisition module, which is used to collect video image data in real time through monitoring cameras distributed in the security area, and perform basic formatting processing on the data; a target detection and trajectory generation module, which is used to extract dynamic targets, distinguish between background static elements and target dynamic features, and generate initial detection data of dynamic targets; based on the spatial correlation of multiple cameras, a multi-view spatial motion model of dynamic targets is constructed; the real-time position of dynamic targets is recorded, and the speed, acceleration and residence time of the targets are calculated in combination with time series information to generate trajectory data; a scene risk assessment and path prediction module, which is used to calculate the cumulative risk value of the target path based on dynamic motion trajectory data and scene layout, and generate preliminary path risk distribution data; the key monitoring features of the region are integrated, the path risk weight is corrected, and the comprehensive risk value is generated; the most likely intrusion path is identified, the path prediction priority is dynamically adjusted, the potential paths are prioritized and the path ranking data is output; a dynamic sensitivity adjustment module, which is used to dynamically adjust the monitoring sensitivity according to the predicted path priority data; a multi-level warning trigger and information push module, which is used to trigger the multi-level warning mechanism and output the alarm level when the path priority exceeds the preset threshold.

[0046] Compared with the prior art, the beneficial effects achieved by the present invention are: by combining technologies such as dynamic target detection, trajectory analysis, path prediction and risk assessment, a more efficient intrusion warning mechanism is realized; the dynamic characteristics of the target are accurately extracted by using a multi-view spatial motion model and a trajectory data correction algorithm; high-risk paths are effectively identified through comprehensive risk assessment and dynamic path sorting; and based on a sensitivity dynamic adjustment strategy and a multi-level warning mechanism, the real-time and intelligence of the monitoring system are significantly improved. The present invention solves the problems of insufficient path prediction accuracy, lack of intelligence in risk assessment, and inflexible adjustment of warning sensitivity in the prior art, and belongs to the category of video image intelligent analysis and warning methods in the field of security technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0048] Figure 1 It is a schematic diagram of the steps of a security area intrusion warning method based on video images of the present invention;

[0049] Figure 2 The present invention is a schematic structural diagram of a security area intrusion warning system based on video images. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] Example 1

[0052] See also Figure 1 In this embodiment 1: a method for early warning of intrusion in a security area based on video images is provided, the method comprising the following steps:

[0053] Step S1: The surveillance cameras distributed in the security area collect video image data in real time and perform basic formatting on the data.

[0054] Step S2: Using a preset target detection algorithm, extract the intrusion target from the video image, record the trajectory of the intrusion target, analyze the movement direction, speed change and residence time of the intrusion target, and generate a dynamic motion trajectory.

[0055] Step S2.1: Based on the basic formatting processing result of step S1, a hierarchical target detection algorithm is used to extract dynamic targets from the video image, distinguish between background static elements and target dynamic features, and generate initial detection data of the dynamic targets, as follows:

[0056]

[0057] in, represents the initial detection data of target i, represents the weight of the nth feature, represents the data value of target i on the nth feature, Indicates the number of target features.

[0058] Step S2.2: Based on the initial detection data , using the spatial correlation of multiple cameras in the area, matching the position changes of dynamic targets in different perspectives, and constructing a multi-perspective spatial motion model of the dynamic target, as follows:

[0059]

[0060] in, is the multi-view spatial motion model of the dynamic target k, is the initial detection data of target k calculated in the cth camera, is the mean of the initial detection data of target k in all cameras, and C is the total number of cameras.

[0061] Step S2.3: Based on the multi-view spatial motion model of the dynamic target, the real-time position of the dynamic target is recorded, and the speed change, acceleration and dwell time of the dynamic target are calculated in combination with the time series information to generate the trajectory data of the dynamic target, as follows:

[0062]

[0063]

[0064] in, is the velocity of target k at time t, is the acceleration of target k at time t, is the position of target k at time t, is the time interval, For target k at time location, For target k at time speed.

[0065] Step S2.4: Based on the trajectory data, match it with the scene feature database of the security area to correct the abnormal points in the trajectory, as follows:

[0066]

[0067] in, is the corrected trajectory point of target k at time t, is the multi-view spatial motion model of the dynamic target k, is the position of target k at time t, and Represents the multi-view spatial motion model of dynamic target k and the position of target k at time t The weight coefficient of .

[0068] Based on the corrected trajectory data, a dynamic motion trajectory dataset is generated, and the trajectory dataset is passed as input to step S3.

[0069] Step S3: Based on the dynamic motion trajectory and the scene layout in the area, the trajectory prediction model is used to infer the intrusion path; the danger level of different potential paths is evaluated, focusing on the paths close to the core area or with abnormal movement directions, and the paths are sorted by priority.

[0070] Step S3.1: Based on the dynamic motion trajectory data and the scene layout data of the security area, the cumulative risk value of the target path passing through different areas is calculated to obtain the preliminary path risk distribution data, as follows:

[0071]

[0072] in, is the initial risk value of the path segment m→n, is the scene feature value of the path segment m→n on feature p, is the length of the path segment m→n, is the weight of feature p, is the normalized coefficient of feature p, is the number of scene features.

[0073] Step S3.2: The preliminary path risk distribution data is integrated with the key monitoring characteristics of the region, the risk weight of the potential path is corrected, and the comprehensive risk value of the path segment m→n is generated as follows:

[0074]

[0075] in, is the comprehensive risk value of the corrected path segment m→n, is the weight coefficient of the key monitoring area, is the key monitoring feature value of the path segment m→n.

[0076] Step S3.3: Using the path prediction model, identify the most likely intrusion path from the regional risk matrix, and dynamically adjust the path prediction priority based on the real-time direction and speed changes of the target movement, as follows:

[0077]

[0078] in, is the priority of the path segment m→n, is the speed of the target at the starting node m of the path segment, is the straight-line distance of path segment m→n.

[0079] Step S3.4: Prioritize all potential paths, focus on paths close to high-risk areas or with abnormal target behaviors, and output path ranking data.

[0080] Step S4: According to the path priority, the early warning sensitivity of the monitoring area is adjusted in real time, such as increasing the detection frequency for high-risk paths or enhancing the detection threshold of the algorithm. When the path analysis results show that the target may be approaching a high-risk area, the early warning system is triggered, an alarm message is issued, and the path data is pushed to the security personnel.

[0081] Step S4.1: According to the priority data of the predicted path, the monitoring sensitivity of the security area is dynamically adjusted as follows:

[0082]

[0083] in, is the monitoring sensitivity after dynamic adjustment, As the basic monitoring sensitivity, is the sensitivity adjustment factor, is the priority of the path segment m→n, The maximum value of all path priorities.

[0084] Step S4.2: For high-priority path segments, the accuracy of the image processing algorithm is increased, and the path priority is updated in real time by further analyzing the subtle features of the target (such as posture changes and speed fluctuations), as follows:

[0085]

[0086] in, is the update weight of the target feature, is the real-time speed fluctuation value of the target, is the target attitude change rate, is the speed fluctuation weight coefficient, is the weight coefficient of posture change.

[0087] Step S4.3: When the path priority exceeds the preset threshold, the multi-level warning mechanism is triggered to output the alarm level and related target data.

[0088] Step S4.4: Push the alarm information and path data to the security personnel in real time, and generate a visual risk map showing the target's real-time location, movement trajectory and predicted path.

[0089] Example 2

[0090] See also Figure 2In this embodiment 2: a security area intrusion warning system based on video images is provided, the system includes: a monitoring video acquisition module, which is used to collect video image data in real time through monitoring cameras distributed in the security area, and perform basic formatting processing on the data; a target detection and trajectory generation module, which is used to extract dynamic targets, distinguish between background static elements and target dynamic features, and generate initial detection data of dynamic targets; based on the spatial correlation of multiple cameras, a multi-view spatial motion model of dynamic targets is constructed; the real-time position of dynamic targets is recorded, and the speed, acceleration and residence time of the targets are calculated in combination with time series information to generate trajectory data; a scene risk assessment and path prediction module, which is used to calculate the cumulative risk value of the target path based on the dynamic motion trajectory data and the scene layout, and generate preliminary path risk distribution data; integrate the key monitoring features of the area, correct the path risk weight, and generate a comprehensive risk value; identify the most likely intrusion path, dynamically adjust the path prediction priority, prioritize the potential paths and output the path ranking data; a dynamic sensitivity adjustment module, which is used to dynamically adjust the monitoring sensitivity according to the predicted path priority data; a multi-level warning trigger and information push module, which is used to trigger the multi-level warning mechanism and output the alarm level when the path priority exceeds the preset threshold.

[0091] Example 3

[0092] In this embodiment 3: a security area intrusion warning method based on video images is provided. In order to verify the beneficial effects of the present invention, a scientific demonstration is carried out through simulation experiments.

[0093] In order to verify the effectiveness of the security area intrusion path prediction method based on dynamic motion trajectory analysis, 12 surveillance cameras were deployed in a security area, and the field of view covered by each camera overlapped to form a complete area coverage. The area was divided into five key areas: entrance area, general area, high-risk area, core area and exit area. Dynamic motion simulation equipment was arranged in the area to simulate different intrusion targets to test the accuracy of path prediction and the effect of sensitivity adjustment.

[0094] During the experiment, the real-time video data collected by the camera was processed using a preset hierarchical target detection algorithm to extract the initial detection data of dynamic targets. The multi-view spatial motion model of dynamic targets was constructed through the spatial correlation of cameras in the area to generate target trajectory data. The trajectory data was matched with the scene feature database, and the final trajectory data set was formed after correcting the abnormal points. Then, the path risk value was calculated based on the trajectory data combined with the scene layout, the potential intrusion path was predicted, and the paths were prioritized.

[0095] In order to verify that real-time sensitivity adjustment can improve the monitoring effect of high-risk areas, the monitoring sensitivity is dynamically adjusted when the path priority exceeds the set threshold. At the same time, the multi-level early warning mechanism is triggered, the path and alarm information are pushed to security personnel, and a visual risk map is generated to assist decision-making.

[0096] Table 1 Experimental data of dynamic target path prediction in security area

[0097] Test subjects Target Number Initial speed (m / s) Acceleration (m / s²) Dwell time(s) Path risk value Path priority Alert Level Target A 1 1.5 0.2 15 0.85 0.90 high Target B 2 2.0 0.1 10 0.72 0.80 middle Objective-C 3 1.2 0.3 20 0.68 0.75 Low Target D 4 1.8 0.2 12 0.88 0.95 high Target E 5 1.6 0.25 18 0.79 0.85 middle Target F 6 2.2 0.15 8 0.92 0.98 high

[0098] As can be seen from Table 1, the method of the present invention can effectively distinguish the trajectory characteristics of different targets and perform reasonable path sorting according to the path risk value and priority. In high-risk areas (such as targets D and F), the path risk value and priority are significantly higher than other areas. By dynamically adjusting the sensitivity, the system can quickly identify high-risk paths and trigger alarms to ensure regional safety.

[0099] Compared with the existing technology, the traditional method only relies on a single camera or static risk assessment, and cannot predict the target path in real time and dynamically adjust the sensitivity. This embodiment improves the accuracy of intrusion target detection and path prediction by introducing a multi-view spatial motion model and trajectory prediction algorithm; at the same time, through the dynamic sensitivity adjustment mechanism, it effectively reduces the false alarm rate and improves the system response speed.

[0100] The present invention has obvious advantages in real-time early warning and risk identification capabilities in high-risk areas, providing a more intelligent and accurate technical means for security management. Through experimental data verification, the present method has significant innovation and practical value in trajectory analysis, path prediction and risk assessment.

[0101] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0102] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A security area intrusion warning method based on video images, characterized in that: The method comprises the following steps: Step S1: Collect video image data in real time through surveillance cameras distributed in the security area, and perform basic formatting on the data; Step S2: using a preset target detection algorithm to extract the intrusion target from the video image, record the trajectory of the intrusion target, analyze the movement direction, speed change and residence time of the intrusion target, and generate a dynamic motion trajectory; Step S3: Based on the dynamic motion trajectory and the scene layout in the area, the trajectory prediction model is used to infer the intrusion path; the danger level of different potential paths is evaluated and the paths are sorted by priority; Step S4: According to the path priority, the warning sensitivity of the monitoring area is adjusted in real time. If the path analysis result is a high-risk area, the warning system is triggered and an alarm message is issued; The specific implementation process of step S2 includes: Step S2.1: Based on the basic formatting processing result of step S1, a hierarchical target detection algorithm is used to extract dynamic targets from the video image, distinguish between background static elements and target dynamic features, and generate initial detection data of the dynamic targets, as follows: Among them, D i represents the initial detection data of target i, ω n represents the weight of the nth feature, d i，n represents the data value of target i on the nth feature, and N represents the number of target features; Step S2.2: Based on the initial detection data D i , using the spatial correlation of multiple cameras in the area, matching the position changes of dynamic targets in different perspectives, and constructing a multi-perspective spatial motion model of the dynamic target, as follows: Among them, M k is the multi-view spatial motion model of dynamic target k, D k,c is the initial detection data of target k calculated in the cth camera, is the mean of the initial detection data of target k in all cameras, and C is the total number of cameras; Step S2.3: Based on the multi-view spatial motion model of the dynamic target, the real-time position of the dynamic target is recorded, and the speed change, acceleration and dwell time of the dynamic target are calculated in combination with the time series information to generate the trajectory data of the dynamic target, as follows: Among them, V k (t) is the speed of target k at time t, A k (t) is the acceleration of target k at time t, P k (t) is the position of target k at time t, Δt is the time interval, P k (t+Δt) is the position of target k at time t+Δt, V k (t+Δk) is the velocity of target k at time t+Δt; Step S2.4: Based on the trajectory data, match it with the scene feature database of the security area to correct the abnormal points in the trajectory, as follows: in, is the corrected trajectory point of target k at time t, M k is the multi-view spatial motion model of dynamic target k, P k (t) is the position of target k at time t, λ1 and λ2 represent the multi-view spatial motion model M of dynamic target k respectively. k and the position P of the target k at time t k The weight coefficient of (t); Generate a dynamic motion trajectory data set based on the corrected trajectory data, and pass the trajectory data set as input to step S3; The specific implementation process of step S3 includes: Step S3.1: Based on the dynamic motion trajectory data and the scene layout data of the security area, the cumulative risk value of the target path passing through different areas is calculated to obtain the preliminary path risk distribution data, as follows: Among them, R mn is the initial risk value of path segment m→n, T mn，p is the scene feature value of path segment m→n on feature p, L mn is the length of the path segment m→n, w p is the weight of feature p, σ p is the normalization coefficient of feature p, P is the number of scene features; Step S3.2: The preliminary path risk distribution data is integrated with the key monitoring characteristics of the region, the risk weight of the potential path is corrected, and the comprehensive risk value of the path segment m→n is generated as follows: in, is the modified comprehensive risk value of the path segment m→n, β is the weight coefficient of the key monitoring area, C mn is the key monitoring feature value of the path segment m→n; Step S3.3: Using the path prediction model, the intrusion path is identified from the regional risk matrix, and the path prediction priority is dynamically adjusted in combination with the real-time direction and speed changes of the target movement, as follows: Among them, P mn is the priority of path segment m→n, S m is the speed of the target at the starting node m of the path segment, D mn is the straight-line distance of path segment m→n; Step S3.4: prioritize all potential paths and output path ranking data; The specific implementation process of step S4 includes: Step S4.1: According to the priority data of the predicted path, the monitoring sensitivity of the security area is dynamically adjusted as follows: Among them, S adj is the monitoring sensitivity after dynamic adjustment, S0 is the basic monitoring sensitivity, γ is the sensitivity adjustment coefficient, P mn is the priority of path segment m→n, and max(P) is the maximum value of all path priorities; Step S4.2: For high priority path segments, the path priority is updated in real time, as follows: W f =α·V f +δ·ΔA Among them, W f is the update weight of the target feature, V f is the real-time speed fluctuation value of the target, ΔA is the target attitude change rate, α is the speed fluctuation weight coefficient, and δ is the attitude change weight coefficient; Step S4.3: When the path priority exceeds the preset threshold, a multi-level warning mechanism is triggered and an alarm is output; Step S4.4: Push the alarm information and path data to the security personnel in real time, and generate a visual risk map showing the target's real-time location, movement trajectory and predicted path.

2. A security area intrusion warning system based on video images, based on the security area intrusion warning method based on video images according to claim 1, characterized in that: The surveillance video acquisition module is used to collect video image data in real time through surveillance cameras distributed in the security area and perform basic formatting processing on the data; The target detection and trajectory generation module is used to extract dynamic targets, distinguish between background static elements and target dynamic features, and generate initial detection data for dynamic targets; build a multi-view spatial motion model for dynamic targets based on the spatial correlation of multiple cameras; record the real-time position of dynamic targets, calculate the target's speed, acceleration and dwell time based on time series information, and generate trajectory data; The scenario risk assessment and path prediction module is used to calculate the cumulative risk value of the target path based on the dynamic motion trajectory data and the scenario layout, and generate preliminary path risk distribution data; Integrate key monitoring features of the region, modify path risk weights, and generate comprehensive risk values; Identify intrusion paths, dynamically adjust path prediction priorities, prioritize potential paths and output path ranking data; A dynamic sensitivity adjustment module is used to dynamically adjust the monitoring sensitivity according to the predicted path priority data; The multi-level warning trigger and information push module is used to trigger the multi-level warning mechanism and output the alarm level when the path priority exceeds the preset threshold.

3. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for early warning of intrusion into a security area based on video images as described in claim 1 are implemented.

4. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a method for early warning of intrusion into a security area based on video images as described in claim 1 are implemented.

Images