Deep learning-based intelligent identification method and system for intelligent marketing and defense monitoring

Abstract

The invention discloses a deep learning-based intelligent marketing and defense monitoring intelligent identification method and system, and belongs to the technical field of monitoring identification. Position information and standard operation processes of all posts in the camp are obtained, the staying duration, the moving path and the moving duration of a target individual in a normal state are collected, and a standard behavior chain of the target individual is constructed; a multi-camera collaborative perception and image recognition technology is utilized to detect a shelter in real time and establish a two-dimensional scatter coordinate system; when the target path intersects with the shielding region, generating a detour candidate region and constructing a replacement path; calculating virtual correction movement duration through path reconstruction, and constructing a virtual behavior prediction chain; and finally, the actual behavior is compared with the virtual path, and abnormal recognition early warning is realized in combination with a preset threshold value, so that the individual behavior recognition precision in a complex shielding environment and the intelligent level of a monitoring system are improved, and the behavior recognition precision in a complex scene and the intelligent level of the monitoring system are improved.

Description

Technical Field

[0001] This invention relates to the field of surveillance and identification technology, specifically to a smart defense surveillance and identification method and system based on deep learning. Background Technology

[0002] With the rapid development of artificial intelligence, image recognition, and deep learning technologies, intelligent monitoring systems are increasingly being used in critical scenarios such as military camps, industrial parks, and disaster relief sites. Traditional camp monitoring systems primarily rely on fixed cameras and manual patrols. Their monitoring capabilities are limited by visual obstructions, personnel fatigue, and dynamic changes in complex environments, making it difficult to achieve real-time identification and accurate early warning of abnormal behavior. Currently, most intelligent behavior recognition systems only perform behavior discrimination through feature extraction and classification models from static images, lacking full-chain modeling of the target individual's behavioral process. Especially when there are temporary obstructions or dynamic changes in the environmental structure, they cannot effectively update the semantic information of the monitoring scene, leading to an increased false positive rate.

[0003] Existing technologies often neglect detailed modeling of standard operating procedures and fail to incorporate the temporal sequence and spatial paths of individual behaviors across different roles as key elements into the monitoring and analysis logic. Furthermore, existing identification systems lack dynamic correction and detour reasoning mechanisms for potential path intersections between target behaviors and obstructions, making it difficult to simulate and predict the potential behavioral paths of individual targets in abnormal states, thus reducing the accuracy and timeliness of identification.

[0004] Existing camp defense and surveillance technologies generally suffer from the following key problems: First, they lack the ability to dynamically model target behavior, especially when faced with occlusion and path changes, they cannot automatically generate virtual corrected paths; second, they cannot integrate image perception, spatial path modeling, and standard behavior chain prediction, and lack a multi-source information fusion processing mechanism; third, early warning mechanisms rely on static rules and cannot be dynamically adjusted based on behavior deviations. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for intelligent identification of smart camp defense monitoring based on deep learning, so as to solve the problems mentioned in the background art.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] This invention relates to a deep learning-based intelligent identification method for camp security monitoring. The method includes the following steps: Step S1: Acquire the location information and standard operating procedures of all positions within the camp, collect the dwell time, movement path, and movement duration of the target individual under normal conditions, and construct the target individual's standard behavior chain; Step S2: Collect real-time monitoring images of the camp, identify temporary obstructions and their location information; construct a two-dimensional scattered coordinate system for the camp based on the location information of all positions within the camp, the location information of temporary obstructions, and the standard behavior chain; Step S3: Based on the camp's two-dimensional scattered coordinate system, if it is determined that a temporary obstruction exists on the movement path, construct a set of intersecting path coordinate points and a detour candidate coordinate region; calibrate key coordinate anchor points and extract the starting and ending intersection coordinate points; Step S4: Based on the starting and ending intersection coordinate points, construct the starting and ending detour coordinate points, generate candidate detour path segments, and calculate the virtual corrected movement duration; Step S5: Construct a virtual behavior prediction chain, monitor the actual continuous behavior data of the target individual in real time; preset thresholds, analyze, and perform intelligent identification and early warning.

[0008] As a preferred embodiment of the intelligent identification method for smart camp defense monitoring based on deep learning described in this invention, the location information and standard operating procedures of all positions within the camp are obtained, and continuous behavioral data of the target individual from the i-th position to the j-th position under normal conditions are collected. The continuous behavioral data includes the target individual's dwell time, movement path, and movement time.

[0009] Based on the continuous behavioral data, a standard behavioral chain from the i-th position to the j-th position is constructed, denoted as SBC. i,j ={P i ,LS i ,MP i,j ,MT i,j ,P j}, where LS i MP represents the duration of time a target individual stays in the i-th position. i,j MT represents the movement path of a target individual from the i-th position to the j-th position. i,j P represents the time it takes for a target individual to move from the i-th position to the j-th position. i Let P represent the i-th position. j This represents the j-th position.

[0010] It should be noted that by collecting location information, standard operating procedures, and the dwell time, movement path, and movement duration of target individuals under normal conditions for all positions within the camp, structured modeling of individual behavior was achieved, constructing a standard behavior chain composed of continuous behavioral data across multiple positions. This establishes a reference paradigm for individual behavior, providing a benchmark for subsequent behavioral deviation identification. This step establishes behavioral patterns through a data-driven approach, without relying on human experience judgment, and possesses the advantages of strong adaptability and high modeling accuracy.

[0011] As a preferred embodiment of the intelligent identification method for smart camp defense monitoring based on deep learning described in this invention, several fixed monitoring cameras are deployed according to the area of ​​the camp, and multi-camera collaborative perception technology is used in the camp defense monitoring system to collect monitoring images of the camp in real time; based on deep learning and image perception technology, temporary obstructions and their location information in the camp are identified in the monitoring images, wherein the temporary obstructions include training equipment piles and vehicles.

[0012] Based on the location information of all positions within the camp, the location information of temporary shelters, and the Standard Behavior Chain (SBC) i,j Construct a two-dimensional scattered coordinate system for the camp, as follows:

[0013] Based on the camp's area, a two-dimensional scattered coordinate system is constructed for the camp, and the Standard Behavior Chain (SBC) is defined. i,j movement path MP i,j The movement path is represented as a sequence of discrete coordinate points in the two-dimensional scattered coordinate system of the camp, denoted as ZMP. i,j ={(x n ,y n )|n∈[1,N]}, where, (x n ,y n ) represents the movement path MP i,j The nth coordinate point in the two-dimensional scattered coordinate system of the camp, where N represents the movement path MP. i,j The total number of coordinate points in the two-dimensional scattered coordinate system of the camp;

[0014] The location information of the temporary shelter is mapped onto the two-dimensional scattered coordinate system of the campsite to form a coordinate region, and the range of the horizontal coordinate of the coordinate region is denoted as [x...]. a,min ,x a,max The range of the ordinates of the coordinate region is denoted as [y]. a,min ,y a,max ], where x a,min Let x represent the minimum x-coordinate of the a-th temporary occlusion. a,max The x-coordinate of the a-th temporary occlusion is the maximum value. a,min Let y represent the minimum ordinate of the a-th temporary occlusion.a,max This represents the maximum value of the ordinate of the a-th temporary occlusion.

[0015] It should be noted that by deploying multiple cameras and integrating deep learning and image perception technologies, the system can identify temporary obstructions such as training equipment piles and vehicles, along with their location information, in real time. This achieves precise spatial mapping of obstruction factors and, together with job positions and standard behavioral chains, constructs a two-dimensional scattered coordinate system for the camp. This allows for the creation of a spatial foundation model that can be used for path analysis, detour decisions, and subsequent behavior prediction. This step significantly enhances the system's adaptability to complex environments, enabling the monitoring system to dynamically respond to changes in spatial structure.

[0016] As a preferred embodiment of the intelligent identification method for smart camp defense monitoring based on deep learning described in this invention, based on the camp's two-dimensional scattered coordinate system, if x a,min ≤x n ≤x a,max , and y a,min ≤y n ≤y a,max Then determine if the a-th temporary occlusion has a movement path MP. i,j The above constructs a virtual behavior prediction chain for the target individual from the i-th position to the j-th position, as follows:

[0017] Construct a discrete coordinate sequence (ZMP) of the movement path of a target individual from position i to position j. i,j The set of intersection path coordinates with the coordinate region of the temporary obstruction is denoted as P. obs ={(x m ,y m )|x a,mim ≤x m ≤x a,max y a,min ≤y m ≤y a,max ,(x m ,y m )∈MP i,j}, where (x m ,y m MP represents the movement path of a target individual from position i to position j. i,j The coordinates of the m-th intersecting path point with the coordinate region of the temporary obstruction;

[0018] A preset extension distance r is used to extend the coordinate region of the temporary obstruction by an additional extension distance r. The extended coordinate region is denoted as the detour candidate coordinate region, and the range of the x-coordinate of the detour candidate coordinate region is denoted as [x...]. a,min -r,x a,max +r], the range of the ordinate of the candidate coordinate region for the detour is denoted as [y]. a,min-r,y a,max +r], and mark the set of key coordinate anchor points on the boundary of the candidate coordinate region of the detour, as follows:

[0019] Based on the x-coordinate range of the candidate coordinate region for detour [x a,min -r,x a,max +r] and the range of the ordinate [y a,min -r,y a,max +r], the set of key coordinate anchor points is denoted as KSA={NW(x a,min -r,y a,max +r),NE(x a,max +r,y a,max +r),SW(x a,min -r,y a,min -r), SE(x a,max +r,y a,min -r)};

[0020] Extract the discrete coordinate sequence (ZMP) of the movement path of the target individual from the i-th position to the j-th position. i,j The first and last coordinate points intersecting with the candidate coordinate region of the detour are denoted as the initial intersection point (x). in ,y in ) and the final intersection point (x out ,y out ).

[0021] It should be noted that by determining whether temporary obstructions intersect with individual standard paths, the intersection points between the path and the obstruction area are extracted, and a detour candidate region is constructed using an extended distance *r*. This achieves geometric identification of path interference areas and boundary delineation of the detour strategy space, thereby providing spatial anchors and reasonable boundaries for intelligently generating obstacle avoidance paths, improving the intelligence and accuracy of path adjustment. This step, through the construction of intersecting path points and key anchors, provides efficient support for path reconstruction while preserving the reference value of the original path structure.

[0022] As a preferred embodiment of the intelligent identification method for smart camp defense monitoring based on deep learning described in this invention, points (x, y, y) that intersect with the initial coordinates are selected from the key coordinate anchor point set KSA. in ,y in ) and the final intersection point (x out ,y out The nearest critical coordinate anchor point is denoted as the starting detour coordinate point KSA. in and the final detour coordinates KSA out ;

[0023] Starting from the coordinate point KSA inStarting from point KSA, the coordinates of the detour end point KSA. out Using the destination as the endpoint, and combining the camp's two-dimensional scattered coordinate system and the candidate detour coordinate region, the A* algorithm is used to generate candidate detour path segments, denoted as R. in-out The discrete coordinate sequence ZMP of the movement path of the target individual from the i-th position to the j-th position. i,j The initial intersection point (x) in ,y in ) to the final intersection point (x out ,y out The coordinate point sequence of ) is replaced with candidate detour path segment R. in-out And marked as the virtual modified movement path (DMP) of the target individual from the i-th position to the j-th position. i,j ;

[0024] Based on Standard Behavior Chain (SBC) i,j ={P i ,LS i ,MP i,j ,MT i,j ,P j} and Virtual Correction Movement Path (DMP) i,j The virtual adjusted movement time for a target individual from position i to position j is calculated using the following formula: Among them, DMT i,j d(DMP) represents the virtual modified movement time of the target individual from the i-th position to the j-th position. i,j ) indicates Virtual Modified Movement Path (DMP) i,j The length of d(MP) i,j ) represents the movement path MP i,j Length, This represents the preset terrain complexity coefficient.

[0025] It should be noted that by selecting the key coordinate anchor point closest to the path intersection as the start and end detour points, and combining this with the A* algorithm to generate candidate detour paths, which then replace the occluded sections in the original path, dynamic correction and simulation generation of individual behavior paths are achieved. This allows for the construction of a "reasonable hypothetical path" for individual behavior under temporary occlusion conditions, and the calculation of virtual corrected movement time based on this. This step not only preserves path continuity and practical feasibility but also incorporates terrain complexity into the movement time correction calculation, improving the rationality and adaptability of spatiotemporal modeling.

[0026] As a preferred embodiment of the intelligent identification method for smart camp defense monitoring based on deep learning described in this invention, the virtual behavior prediction chain is denoted as DSBC. i,j ={P i ,LS i DMPi,j DMT i,j ,P j}, monitor the actual continuous behavior data of the target individual in real time, and preset the movement path threshold and movement duration threshold;

[0027] If the actual movement path of the target individual differs from the virtual corrected movement path DMP i,j The difference between the actual movement time and the virtual corrected movement time (DMT) of the target individual is greater than or equal to the movement path threshold, and the difference between the actual movement time and the virtual corrected movement time (DMT) of the target individual is greater than or equal to the movement path threshold. i,j If the difference is greater than or equal to the movement time threshold, it is determined that the target individual has an abnormal movement path and movement time from the i-th position to the j-th position, and an early warning is issued to the relevant staff.

[0028] The system updates the two-dimensional scatter coordinate system of the camp and the coordinate areas of temporary obstructions in real time, enabling intelligent identification and early warning for camp defense monitoring.

[0029] It should be noted that by comparing the corrected virtual behavior chain with the real-time behavior of the target individual and setting thresholds to determine whether the behavior path and time deviate from the standard, a rapid detection and system early warning mechanism for abnormal behavior of the target individual is achieved. This enables intelligent and automated security monitoring and intervention in cases of temporary obstruction, non-preset events, or emergencies. This step significantly improves the response time and recognition accuracy of the monitoring system, effectively preventing misjudgments of behavior caused by viewpoint obstruction or non-standard paths.

[0030] The intelligent identification system for smart camp defense monitoring based on deep learning includes: a behavior baseline modeling module, a dynamic scene coordinate construction module, a path conflict area calibration module, a detour path planning and correction module, and a behavior anomaly early warning and monitoring module.

[0031] The behavior baseline modeling module acquires the location information and standard operating procedures of all positions within the camp, collects the dwell time, movement path and movement duration of the target individual under normal conditions, and constructs the standard behavior chain of the target individual.

[0032] The dynamic scene coordinate construction module: collects real-time monitoring images of the camp, identifies temporary obstructions and their location information; and constructs a two-dimensional scattered coordinate system for the camp based on the location information of all positions in the camp, the location information of temporary obstructions, and the standard behavior chain.

[0033] The path conflict area identification module: based on the camp's two-dimensional scattered coordinate system, if it is determined that a temporary obstruction exists on the movement path, it constructs a set of intersecting path coordinate points and a detour candidate coordinate area; it calibrates key coordinate anchor points and extracts the starting and ending intersection coordinate points;

[0034] The detour path planning and correction module: Based on the starting intersection coordinate point and the ending intersection coordinate point, it constructs the starting detour coordinate point and the ending detour coordinate point, generates candidate detour path segments, and calculates the virtual correction movement time;

[0035] The abnormal behavior early warning and monitoring module: constructs a virtual behavior prediction chain to monitor the actual continuous behavior data of the target individual in real time; presets thresholds, analyzes and performs intelligent identification and early warning.

[0036] Furthermore, the dynamic scene coordinate construction module includes an occlusion acquisition unit and a coordinate system construction unit;

[0037] The obstruction acquisition unit: Based on the area of ​​the camp, deploy several fixed monitoring cameras, and use multi-camera collaborative perception technology in the camp defense monitoring system to collect monitoring images of the camp in real time; based on deep learning and image perception technology, identify temporary obstructions in the camp and their location information in the monitoring images, wherein the temporary obstructions include training equipment piles and vehicles.

[0038] The coordinate system construction unit is based on the location information of all positions within the camp, the location information of temporary cover, and the Standard Behavior Chain (SBC). i,j Construct a two-dimensional scattered coordinate system for the campsite, as follows: Based on the campsite's area, construct a two-dimensional scattered coordinate system for the campsite, and define the Standard Behavior Chain (SBC). i,j movement path MP i,j The movement path is represented as a sequence of discrete coordinate points in the camp's two-dimensional scattered coordinate system; the location information of the temporary obstruction is mapped to the camp's two-dimensional scattered coordinate system to form a coordinate region, and the range of the x-coordinate of the coordinate region is denoted as [x...]. a,min ,x a,max The range of the ordinates of the coordinate region is denoted as [y]. a,min ,y a,max ].

[0039] Furthermore, the path conflict region identification module includes a virtual behavior prediction chain construction unit and an identification unit;

[0040] The virtual behavior prediction chain construction unit: based on the camp's two-dimensional scattered coordinate system, if x a,min ≤x n ≤x a,max , and y a,min ≤y n ≤y a,max Then determine if the a-th temporary occlusion has a movement path MP. i,jThe virtual behavior prediction chain of the target individual from the i-th position to the j-th position is constructed as follows: the set of intersection path coordinate points of the discrete coordinate point sequence of the target individual's movement path from the i-th position to the j-th position and the coordinate region of the temporary obstruction is constructed.

[0041] The calibration unit: A preset extension distance r is used to add an extension distance r to the coordinate region of the temporary obstruction. The extended coordinate region is recorded as a detour candidate coordinate region. A set of key coordinate anchor points is calibrated on the boundary of the detour candidate coordinate region, specifically as follows: Based on the horizontal and vertical coordinate ranges of the detour candidate coordinate region; the first and last coordinate points intersecting the discrete coordinate point sequence of the target individual's movement path from the i-th position to the j-th position with the detour candidate coordinate region are extracted and recorded as the initial intersection coordinate points (x, y, y). in ,y in ) and the final intersection point (x out ,y out ).

[0042] Furthermore, the detour path planning and correction module includes a detour path planning unit and a correction unit;

[0043] The detour path planning unit: selects from the key coordinate anchor point set KSA the coordinate points (x) that intersect with the starting point. in ,y in ) and the final intersection point (x out ,y out The nearest critical coordinate anchor point is denoted as the starting detour coordinate point KSA. in and the final detour coordinates KSA out ;

[0044] Starting from the coordinate point KSA in Starting from point KSA, the coordinates of the detour end point KSA. out Using the destination as the endpoint, and combining the camp's two-dimensional scattered coordinate system and the candidate detour coordinate region, the A* algorithm is used to generate candidate detour path segments, denoted as R. in-out The discrete coordinate sequence ZMP of the movement path of the target individual from the i-th position to the j-th position. i,j The initial intersection point (x) in ,y in ) to the final intersection point (x out ,y out The coordinate point sequence of ) is replaced with candidate detour path segment R. in-out And marked as the virtual modified movement path (DMP) of the target individual from the i-th position to the j-th position. i,j ;

[0045] The correction unit is based on standard behavior chains and virtual corrected movement paths (DMP). i,j Calculate the virtual corrected movement time of the target individual from the i-th position to the j-th position.

[0046] Compared with existing technologies, the beneficial effects achieved by this invention are as follows: The intelligent camp defense monitoring and identification method and system based on deep learning provided by this invention acquires the location information and standard operating procedures of all positions in the camp, collects continuous behavioral data of individuals under normal conditions, constructs a standard behavioral chain, and achieves structured modeling of individual behavioral patterns, providing a precise benchmark for subsequent anomaly identification; through multi-camera collaborative perception and deep learning image recognition, it acquires temporary obstructions and their location information in real time, and constructs a two-dimensional scattered coordinate system of the camp, realizing a digital representation of the spatial scene and providing spatial support for path analysis; furthermore, by judging obstructions… By analyzing the intersection relationship with the standard path, extracting the intersection path points, and setting buffer distances to construct detour candidate areas, potential interference with the behavioral path is identified. Alternative path segments are generated based on key coordinate anchor points. The A* algorithm is used to generate detour path fragments, replacing occluded sections and correcting movement duration, achieving dynamic adaptation and reasonable repair of the standard behavioral chain. Finally, by constructing a virtual behavior prediction chain, the actual individual behavior is compared with the corrected path, and a threshold is set to achieve abnormal behavior early warning. This improves the accuracy of individual behavior recognition and the intelligence level of the monitoring system in complex occlusion environments, demonstrating good real-time performance, adaptability, and wide application value. Attached Figure Description

[0047] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0048] Figure 1 This is a schematic diagram illustrating the steps of the intelligent identification method for smart camp defense monitoring based on deep learning in this invention;

[0049] Figure 2 This is a schematic diagram of the intelligent identification system for smart camp defense monitoring based on deep learning, which is the subject of this invention. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] Please see Figure 1In this first embodiment: a smart defense monitoring and intelligent identification method based on deep learning is provided, which includes the following steps:

[0052] Step S1: Obtain the location information and standard operating procedures of all positions in the camp, collect the dwell time, movement path and movement time of the target individual under normal conditions, and construct the standard behavior chain of the target individual.

[0053] Specifically, the location information and standard operating procedures of all positions within the camp are obtained, and continuous behavioral data of the target individual from the i-th position to the j-th position under normal conditions are collected. The continuous behavioral data includes the target individual's dwell time, movement path, and movement time.

[0054] Based on the continuous behavioral data, a standard behavioral chain from the i-th position to the j-th position is constructed, denoted as SBC. i,j ={P i ,LS i ,MP i,j ,MT i,j ,P j}, where LS i MP represents the duration of time a target individual stays in the i-th position. i,j MT represents the movement path of a target individual from the i-th position to the j-th position. i,j P represents the time it takes for a target individual to move from the i-th position to the j-th position. i Let P represent the i-th position. j This represents the j-th position.

[0055] In this invention, traditional monitoring relies solely on manually set static rules (such as prohibiting entry into a certain area). However, normal behavior in a camp setting is highly procedural (such as working in the order of job positions), lacking precise definitions of dynamic behavior between positions. This can easily lead to reasonable deviations being misjudged as abnormal. Therefore, this step collects data on the duration of a target individual's stay at their post, the movement path between posts, and the duration of movement, constructing a standard behavior chain covering all positions. This transforms normal behavior into a quantifiable and comparable digital model, providing a behavioral baseline for the monitoring system. For example, in a routine patrol from sentry post A to equipment depot B in a camp, the standard behavior chain records the complete process of "staying at sentry post A for 5 minutes → moving along the main road for 8 minutes → arriving at equipment depot B," serving as a reference for judging whether subsequent behavior is abnormal.

[0056] Step S2: Collect real-time monitoring images of the camp, identify temporary obstructions and their location information; construct a two-dimensional scattered coordinate system for the camp based on the location information of all positions in the camp, the location information of temporary obstructions, and the standard behavior chain.

[0057] Specifically, based on the area of ​​the camp, several fixed surveillance cameras are deployed, and multi-camera collaborative sensing technology is used in the camp security monitoring system to collect real-time monitoring images of the camp; based on deep learning and image perception technology, temporary obstructions in the camp and their location information are identified in the monitoring images, including training equipment piles and vehicles.

[0058] Based on the location information of all positions within the camp, the location information of temporary shelters, and the Standard Behavior Chain (SBC) i,j Construct a two-dimensional scattered coordinate system for the camp, as follows:

[0059] Based on the camp's area, a two-dimensional scattered coordinate system is constructed for the camp, and the Standard Behavior Chain (SBC) is defined. i,j movement path MP i,j The movement path is represented as a sequence of discrete coordinate points in the two-dimensional scattered coordinate system of the camp, denoted as ZMP. i,j ={(x n ,y n )|n∈[1,N]}, where, (x n ,y n ) represents the movement path MP i,j The nth coordinate point in the two-dimensional scattered coordinate system of the camp, where N represents the movement path MP. i,j The total number of coordinate points in the two-dimensional scattered coordinate system of the camp;

[0060] The location information of the temporary shelter is mapped onto the two-dimensional scattered coordinate system of the campsite to form a coordinate region, and the range of the horizontal coordinate of the coordinate region is denoted as [x...]. a,min ,x a,max The range of the ordinates of the coordinate region is denoted as [y]. a,min ,y a,max ], where x a,min Let x represent the minimum x-coordinate of the a-th temporary occlusion. a,max The x-coordinate of the a-th temporary occlusion is the maximum value. a,min Let y represent the minimum ordinate of the a-th temporary occlusion. a,max This represents the maximum value of the ordinate of the a-th temporary occlusion.

[0061] In this invention, traditional monitoring relies on manual visual observation, which cannot quantify the spatial relationship between obstructions and paths. This makes it impossible for computers to automatically determine whether obstructions affect paths and the extent of their impact, thus interrupting the behavior recognition process. Therefore, this step uses multi-camera collaborative perception and deep learning image recognition to capture the location of temporary obstructions in real time. It integrates elements such as camp posts, standard paths, and obstructions into a two-dimensional scatter coordinate system to achieve a digital twin of the physical scene, transforming camp spatial information into a computer-analyzable digital map. For example, when a vehicle temporarily stops on a patrol path, the system marks the rectangular area of ​​the vehicle in the coordinate system (horizontal and vertical coordinate ranges) and clearly presents the spatial relationship between the vehicle and the original path.

[0062] Step S3: Based on the camp's two-dimensional scattered coordinate system, if it is determined that a temporary obstruction exists on the movement path, construct a set of intersecting path coordinate points and a candidate detour coordinate region; calibrate key coordinate anchor points and extract the starting and ending intersecting coordinate points.

[0063] Specifically, based on the two-dimensional scattered coordinate system of the camp, if x a,min ≤x n ≤x a,max , and y a,min ≤y n ≤y a,max Then determine if the a-th temporary occlusion has a movement path MP. i,j The above constructs a virtual behavior prediction chain for the target individual from the i-th position to the j-th position, as follows:

[0064] Construct a discrete coordinate sequence (ZMP) of the movement path of a target individual from position i to position j. i,j The set of intersection path coordinates with the coordinate region of the temporary obstruction is denoted as P. obs ={(x m ,y m )|x a,min ≤x m ≤x a,max y a,min ≤y m ≤y a,max ,(x m ,y m )∈MP i,j}, where (x m ,y m MP represents the movement path of a target individual from position i to position j. i,j The coordinates of the m-th intersecting path point with the coordinate region of the temporary obstruction;

[0065] A preset extension distance r is used to extend the coordinate region of the temporary obstruction by an additional extension distance r. The extended coordinate region is denoted as the detour candidate coordinate region, and the range of the x-coordinate of the detour candidate coordinate region is denoted as [x...]. a,min -r,x a,max +r], the range of the ordinate of the candidate coordinate region for the detour is denoted as [y]. a,min -r,y a,max +r], and mark the set of key coordinate anchor points on the boundary of the candidate coordinate region of the detour, as follows:

[0066] Based on the x-coordinate range of the candidate coordinate region for detour [x a,min -r,x a,max +r] and the range of the ordinate [y a,min -r,y a,max +r], the set of key coordinate anchor points is denoted as KSA={NW(x a,min -r,y a,max +r),NE(x a,max +r,y a,max +r),SW(x a,min -r,y a,min -r), SE(x a,max +r,y a,min -r)};

[0067] Extract the discrete coordinate sequence (ZMP) of the movement path of the target individual from the i-th position to the j-th position. i,j The first and last coordinate points intersecting with the candidate coordinate region of the detour are denoted as the initial intersection point (x). in ,y in ) and the final intersection point (x out ,y out ).

[0068] In this invention, if the intersection range of the obstruction and the path cannot be clearly defined, the subsequent detour path planning will lose its precise spatial boundary, potentially leading to detour paths that are too close (to the obstruction) or too far (increasing invalid distance), affecting the rationality of the expected behavior. Therefore, this step is based on a two-dimensional coordinate system, using coordinate calculations to identify the intersection area between the standard path and the obstruction, extracting the starting intersection point when the path enters the obstruction area and the ending intersection point when it leaves the obstruction area, and delineating detour candidate areas and key anchor points to clarify the interference range of the obstruction on the path, thus accurately locking the interference boundary of the obstruction on the normal path. For example, when a pile of training equipment blocks the patrol path, the system will determine "from which point the personnel enter the obstruction area" and "from which point they leave the obstruction area," and expand the detour candidate area by r meters around the equipment pile, marking four anchor points—northwest, northeast, southwest, and southeast—as detour reference points.

[0069] Step S4: Based on the starting intersection coordinate point and the ending intersection coordinate point, construct the starting detour coordinate point and the ending detour coordinate point, generate candidate detour path segments, and calculate the virtual corrected movement time.

[0070] Specifically, from the key coordinate anchor point set KSA, select the coordinate points (x, y, y) that intersect with the initial coordinates. in ,y in ) and the final intersection point (x out ,y out The nearest critical coordinate anchor point is denoted as the starting detour coordinate point KSA. in and the final detour coordinates KSA out ;

[0071] Starting from the coordinate point KSA in Starting from point KSA, the coordinates of the detour end point KSA. out Using the destination as the endpoint, and combining the camp's two-dimensional scattered coordinate system and the candidate detour coordinate region, the A* algorithm is used to generate candidate detour path segments, denoted as R. in-out The discrete coordinate sequence ZMP of the movement path of the target individual from the i-th position to the j-th position. i,j The initial intersection point (x) in ,y in ) to the final intersection point (x out ,y out The coordinate point sequence of ) is replaced with candidate detour path segment R. in-out And marked as the virtual modified movement path (DMP) of the target individual from the i-th position to the j-th position. i,j ;

[0072] Based on Standard Behavior Chain (SBC) i,j ={P i ,LS i ,MP i,j ,MT i,j ,P j} and Virtual Correction Movement Path (DMP) i,j The virtual adjusted movement time for a target individual from position i to position j is calculated using the following formula: Among them, DMT i,j d(DMP) represents the virtual modified movement time of the target individual from the i-th position to the j-th position. i,j ) indicates Virtual Modified Movement Path (DMP) i,j The length of d(MP) i,j ) represents the movement path MP i,j Length, This represents the preset terrain complexity coefficient.

[0073] In this invention, since occlusion can render the original path invalid, using the original path and duration as the judgment criteria would misjudge deviations in path and duration caused by detours as abnormal behavior, significantly increasing the false alarm rate. Therefore, this step generates an optimal detour path based on the starting and ending intersection points and key anchor points of the conflict area. The movement time is then adjusted according to the detour path length, the original path length, and terrain complexity, constructing a virtual corrected path and virtual corrected duration adapted to the occlusion scenario, providing reasonable behavioral expectations for the occlusion scenario. For example, if the original path is blocked by a vehicle, the system plans a detour path from the east side of the vehicle and calculates that the movement time after the detour should be 10 minutes (original time 8 minutes, but due to the longer path and the grassy surface, the terrain coefficient is taken as 1.25), rather than directly using the original time as the judgment criterion.

[0074] By correcting both the path and duration, the system can reasonably anticipate normal behavior in occluded scenarios, avoiding misjudgments. Furthermore, the A* algorithm is used to generate detour paths, ensuring that the detour route is the shortest and safest, closely resembling the natural detour habits in actual operations.

[0075] Step S5: Construct a virtual behavior prediction chain to monitor the actual continuous behavior data of the target individual in real time; preset thresholds, analyze and perform intelligent identification and early warning.

[0076] Specifically, the virtual behavior prediction chain is denoted as DSBC. i,j ={P i ,LS i DMP i,j DMT i,j ,P j}, monitor the actual continuous behavior data of the target individual in real time, and preset the movement path threshold and movement duration threshold;

[0077] If the actual movement path of the target individual differs from the virtual corrected movement path DMP i,j The difference between the actual movement time and the virtual corrected movement time (DMT) of the target individual is greater than or equal to the movement path threshold, and the difference between the actual movement time and the virtual corrected movement time (DMT) of the target individual is greater than or equal to the movement path threshold. i,j If the difference is greater than or equal to the movement time threshold, it is determined that the target individual has an abnormal movement path and movement time from the i-th position to the j-th position, and an early warning is issued to the relevant staff.

[0078] The system updates the two-dimensional scatter coordinate system of the camp and the coordinate areas of temporary obstructions in real time, enabling intelligent identification and early warning for camp defense monitoring.

[0079] In this invention, this step accurately distinguishes between reasonable deviations and actual anomalies. For example, if the virtual prediction chain is expected to take 10 minutes and the path deviation threshold is 5 meters, when the actual duration is 18 minutes and the path deviates from the predicted path by 15 meters, the system determines it to be an anomaly (there may be personnel leaving their posts, getting lost, etc.) and immediately sends an alert to the monitoring center.

[0080] Please see Figure 2 In this second embodiment: a smart camp defense monitoring and intelligent identification system based on deep learning is provided. The system includes: a behavior baseline modeling module, a dynamic scene coordinate construction module, a path conflict area calibration module, a detour path planning and correction module, and a behavior anomaly early warning and monitoring module.

[0081] The behavior baseline modeling module acquires the location information and standard operating procedures of all positions within the camp, collects the dwell time, movement path and movement duration of the target individual under normal conditions, and constructs the standard behavior chain of the target individual.

[0082] The dynamic scene coordinate construction module: collects real-time monitoring images of the camp, identifies temporary obstructions and their location information; and constructs a two-dimensional scattered coordinate system for the camp based on the location information of all positions in the camp, the location information of temporary obstructions, and the standard behavior chain.

[0083] The path conflict area identification module: based on the camp's two-dimensional scattered coordinate system, if it is determined that a temporary obstruction exists on the movement path, it constructs a set of intersecting path coordinate points and a detour candidate coordinate area; it calibrates key coordinate anchor points and extracts the starting and ending intersection coordinate points;

[0084] The detour path planning and correction module: Based on the starting intersection coordinate point and the ending intersection coordinate point, it constructs the starting detour coordinate point and the ending detour coordinate point, generates candidate detour path segments, and calculates the virtual correction movement time;

[0085] The abnormal behavior early warning and monitoring module: constructs a virtual behavior prediction chain to monitor the actual continuous behavior data of the target individual in real time; presets thresholds, analyzes and performs intelligent identification and early warning.

[0086] Furthermore, the dynamic scene coordinate construction module includes an occlusion acquisition unit and a coordinate system construction unit;

[0087] The obstruction acquisition unit: Based on the area of ​​the camp, deploy several fixed monitoring cameras, and use multi-camera collaborative perception technology in the camp defense monitoring system to collect monitoring images of the camp in real time; based on deep learning and image perception technology, identify temporary obstructions in the camp and their location information in the monitoring images, wherein the temporary obstructions include training equipment piles and vehicles.

[0088] The coordinate system construction unit is based on the location information of all positions within the camp, the location information of temporary cover, and the Standard Behavior Chain (SBC). i,j Construct a two-dimensional scattered coordinate system for the campsite, as follows: Based on the campsite's area, construct a two-dimensional scattered coordinate system for the campsite, and define the Standard Behavior Chain (SBC). i,jmovement path MP i,j The movement path is represented as a sequence of discrete coordinate points in the camp's two-dimensional scattered coordinate system; the location information of the temporary obstruction is mapped to the camp's two-dimensional scattered coordinate system to form a coordinate region, and the range of the x-coordinate of the coordinate region is denoted as [x...]. a,min ,x a,max The range of the ordinates of the coordinate region is denoted as [y]. a,min ,y a,max ].

[0089] Furthermore, the path conflict region identification module includes a virtual behavior prediction chain construction unit and an identification unit;

[0090] The virtual behavior prediction chain construction unit: based on the camp's two-dimensional scattered coordinate system, if x a,min ≤x n ≤x a,max , and y a,min ≤y n ≤y a,max Then determine if the a-th temporary occlusion has a movement path MP. i,j The virtual behavior prediction chain of the target individual from the i-th position to the j-th position is constructed as follows: the set of intersection path coordinate points of the discrete coordinate point sequence of the target individual's movement path from the i-th position to the j-th position and the coordinate region of the temporary obstruction is constructed.

[0091] The calibration unit: A preset extension distance r is used to add an extension distance r to the coordinate region of the temporary obstruction. The extended coordinate region is recorded as a detour candidate coordinate region. A set of key coordinate anchor points is calibrated on the boundary of the detour candidate coordinate region, specifically as follows: Based on the horizontal and vertical coordinate ranges of the detour candidate coordinate region; the first and last coordinate points intersecting the discrete coordinate point sequence of the target individual's movement path from the i-th position to the j-th position with the detour candidate coordinate region are extracted and recorded as the initial intersection coordinate points (x, y, y). in ,y in ) and the final intersection point (x out ,y out ).

[0092] Furthermore, the detour path planning and correction module includes a detour path planning unit and a correction unit;

[0093] The detour path planning unit: selects from the key coordinate anchor point set KSA the coordinate points (x) that intersect with the starting point. in ,y in ) and the final intersection point (x out ,y outThe nearest critical coordinate anchor point is denoted as the starting detour coordinate point KSA. in and the final detour coordinates KSA out ;

[0094] Starting from the coordinate point KSA in Starting from point KSA, the coordinates of the detour end point KSA. out Using the destination as the endpoint, and combining the camp's two-dimensional scattered coordinate system and the candidate detour coordinate region, the A* algorithm is used to generate candidate detour path segments, denoted as R. in-out The discrete coordinate sequence ZMP of the movement path of the target individual from the i-th position to the j-th position. i,j The initial intersection point (x) in ,y in ) to the final intersection point (x out ,y out The coordinate point sequence of ) is replaced with candidate detour path segment R. in-out And marked as the virtual modified movement path (DMP) of the target individual from the i-th position to the j-th position. i,j ;

[0095] The correction unit is based on standard behavior chains and virtual corrected movement paths (DMP). i,j Calculate the virtual corrected movement time of the target individual from the i-th position to the j-th position.

[0096] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0097] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A deep learning-based intelligent identification method for smart camp defense monitoring, characterized in that, The method includes the following steps: Step S1: Obtain the location information and standard operating procedures of all positions in the camp, collect the dwell time, movement path and movement time of the target individual under normal conditions, and construct the standard behavior chain of the target individual; Step S2: Collect real-time monitoring images of the camp, identify temporary obstructions and their location information; construct a two-dimensional scattered coordinate system for the camp based on the location information of all posts in the camp, the location information of temporary obstructions, and the standard behavior chain; Step S3: Based on the camp's two-dimensional scattered coordinate system, if it is determined that a temporary obstruction exists on the movement path, construct a set of intersecting path coordinate points and a candidate detour coordinate region; calibrate key coordinate anchor points and extract the starting and ending intersection coordinate points; Step S4: Based on the starting intersection coordinate point and the ending intersection coordinate point, construct the starting detour coordinate point and the ending detour coordinate point, generate candidate detour path segments and calculate the virtual corrected movement time; Step S5: Construct a virtual behavior prediction chain to monitor the actual continuous behavior data of the target individual in real time; preset thresholds, analyze and perform intelligent identification and early warning.

2. The intelligent identification method for smart camp defense monitoring based on deep learning according to claim 1, characterized in that, The specific implementation process of step S1 includes: Obtain the location information and standard operating procedures of all positions within the camp, and collect continuous behavioral data of the target individual from the i-th position to the j-th position under normal conditions. The continuous behavioral data includes the target individual's dwell time, movement path, and movement duration. Based on the continuous behavioral data, a standard behavioral chain from the i-th position to the j-th position is constructed, denoted as SBC. i,j ={P i ,LS i ,MP i,j ,MT i,j ,P j }, where LS i MP represents the duration of time a target individual stays in the i-th position. i,j MT represents the movement path of a target individual from the i-th position to the j-th position. i,j P represents the time it takes for a target individual to move from the i-th position to the j-th position. i Let P represent the i-th position. j This represents the j-th position.

3. The intelligent identification method for smart camp defense monitoring based on deep learning according to claim 2, characterized in that, The specific implementation process of step S2 includes: Based on the area of ​​the camp, several fixed surveillance cameras are deployed, and multi-camera collaborative sensing technology is used in the camp security monitoring system to collect real-time monitoring images of the camp; based on deep learning and image perception technology, temporary obstructions and their location information in the monitoring images are identified, including training equipment piles and vehicles. Based on the location information of all positions within the camp, the location information of temporary shelters, and the Standard Behavior Chain (SBC) i,j Construct a two-dimensional scattered coordinate system for the camp, as follows: Based on the camp's area, a two-dimensional scattered coordinate system is constructed for the camp, and the Standard Behavior Chain (SBC) is defined. i,j movement path MP i,j The movement path is represented as a sequence of discrete coordinate points in the two-dimensional scattered coordinate system of the camp, denoted as ZMP. i,j ={(x n ,y n )|n∈[1,N]}, where, (x n ,y n ) represents the movement path MP i,j The nth coordinate point in the two-dimensional scattered coordinate system of the camp, where N represents the movement path MP. i,j The total number of coordinate points in the two-dimensional scattered coordinate system of the camp; The location information of the temporary shelter is mapped onto the two-dimensional scattered coordinate system of the campsite to form a coordinate region, and the range of the horizontal coordinate of the coordinate region is denoted as [x...]. a,min ,x a,max The range of the ordinates of the coordinate region is denoted as [y]. a,min ,y a,max ], where x a,min Let x represent the minimum x-coordinate of the a-th temporary occlusion. a,max The x-coordinate of the a-th temporary occlusion is the maximum value. a,min Let y represent the minimum ordinate of the a-th temporary occlusion. a,max This represents the maximum value of the ordinate of the a-th temporary occlusion.

4. The intelligent identification method for smart camp defense monitoring based on deep learning according to claim 3, characterized in that, The specific implementation process of step S3 includes: Based on the two-dimensional scattered coordinate system of the camp, if x a,min ≤x n ≤x a,max , and y a,min ≤y n ≤y a,max Then determine if the a-th temporary occlusion has a movement path MP. i,j The above constructs a virtual behavior prediction chain for the target individual from the i-th position to the j-th position, as follows: Construct a discrete coordinate sequence (ZMP) of the movement path of a target individual from position i to position j. i,j The set of intersection path coordinates with the coordinate region of the temporary obstruction is denoted as P. obs ={(x m ,y m )|x a,min ≤x m ≤x a,max y a,min ≤y m ≤y a,max ,(x m ,y m )∈MP i,j }, where (x m ,y m MP represents the movement path of a target individual from position i to position j. i,j The coordinates of the m-th intersecting path with the coordinate region of the temporary obstruction; A preset extension distance r is used to extend the coordinate region of the temporary obstruction by an additional extension distance r. The extended coordinate region is denoted as the detour candidate coordinate region, and the range of the x-coordinate of the detour candidate coordinate region is denoted as [x...]. a,min -r,x a,max +r], the range of the ordinate of the candidate coordinate region for the detour is denoted as [y]. a,min -r,y a,max +r], and mark the set of key coordinate anchor points on the boundary of the candidate coordinate region of the detour, as follows: Based on the x-coordinate range of the candidate coordinate region for detour [x a,min -r,x a,max +r] and the range of the ordinate [y a,min -r,y a,max +r], the set of key coordinate anchor points is denoted as KSA={NW(x a,min -r,y a,max +r),NE(x a,max +r,y a,max +r),SW(x a,min -r,y a,min -r), SE(x a,max +r,y a,min -r)}; Extract the discrete coordinate sequence (ZMP) of the movement path of the target individual from the i-th position to the j-th position. i,j The first and last coordinate points intersecting with the candidate coordinate region of the detour are denoted as the initial intersection point (x). in ,y in ) and the final intersection point (x out ,y out ).

5. The intelligent identification method for smart camp defense monitoring based on deep learning according to claim 4, characterized in that, The specific implementation process of step S4 includes: Select the coordinate points (x, y) that intersect with the initial coordinates from the key coordinate anchor point set KSA. in ,y in ) and the final intersection point (x out ,y out The nearest critical coordinate anchor point is denoted as the starting detour coordinate point KSA. in and the final detour coordinates KSA out ; Starting from the coordinate point KSA in Starting from point KSA, the coordinates of the detour end point KSA. out Using the destination as the endpoint, and combining the camp's two-dimensional scattered coordinate system and the candidate detour coordinate region, the A* algorithm is used to generate candidate detour path segments, denoted as R. in-out The discrete coordinate sequence ZMP of the movement path of the target individual from the i-th position to the j-th position. i,j The initial intersection point (x) in ,y in ) to the final intersection point (x out ,y out The coordinate point sequence of ) is replaced with candidate detour path segment R. in-out And marked as the virtual modified movement path (DMP) of the target individual from the i-th position to the j-th position. i,j ; Based on Standard Behavior Chain (SBC) i,j ={P i ,LS i ,MP i,j ,MT i,j ,P j } and Virtual Correction Movement Path (DMP) i,j The virtual adjusted movement time for a target individual from position i to position j is calculated using the following formula: Among them, DMT i,j d(DMP) represents the virtual modified movement time of the target individual from the i-th position to the j-th position. i,j ) indicates Virtual Modified Movement Path (DMP) i,j The length of d(MP) i,j ) represents the movement path MP i,j Length, This represents the preset terrain complexity coefficient.

6. The intelligent identification method for smart camp defense monitoring based on deep learning according to claim 5, characterized in that, The specific implementation process of step S5 includes: The virtual behavior prediction chain is DSBC i,j ={P i ,LS i DMP i,j DMT i,j ,P j }, monitor the actual continuous behavior data of the target individual in real time, and preset the movement path threshold and movement duration threshold; If the actual movement path of the target individual differs from the virtual corrected movement path DMP i,j The difference between the actual movement time and the virtual corrected movement time (DMT) of the target individual is greater than or equal to the movement path threshold, and the difference between the actual movement time and the virtual corrected movement time (DMT) of the target individual is greater than or equal to the movement path threshold. i,j If the difference is greater than or equal to the movement time threshold, it is determined that the target individual has an abnormal movement path and movement time from the i-th position to the j-th position, and an early warning is issued to the relevant staff. The system updates the two-dimensional scatter coordinate system of the camp and the coordinate areas of temporary obstructions in real time, enabling intelligent identification and early warning for camp defense monitoring.

7. A deep learning-based intelligent camp defense monitoring and intelligent identification system, executing the deep learning-based intelligent camp defense monitoring and intelligent identification method as described in any one of claims 1-6, characterized in that, The system includes: a behavior baseline modeling module, a dynamic scene coordinate construction module, a path conflict area calibration module, a detour path planning and correction module, and a behavior anomaly early warning and monitoring module. The behavior baseline modeling module acquires the location information and standard operating procedures of all positions within the camp, collects the dwell time, movement path and movement duration of the target individual under normal conditions, and constructs the standard behavior chain of the target individual. The dynamic scene coordinate construction module: collects real-time monitoring images of the camp, identifies temporary obstructions and their location information; and constructs a two-dimensional scattered coordinate system for the camp based on the location information of all positions in the camp, the location information of temporary obstructions, and the standard behavior chain. The path conflict area identification module: based on the camp's two-dimensional scattered coordinate system, if it is determined that a temporary obstruction exists on the movement path, it constructs a set of intersecting path coordinate points and a detour candidate coordinate area; it calibrates key coordinate anchor points and extracts the starting and ending intersection coordinate points; The detour path planning and correction module: Based on the starting intersection coordinate point and the ending intersection coordinate point, it constructs the starting detour coordinate point and the ending detour coordinate point, generates candidate detour path segments, and calculates the virtual correction movement time; The abnormal behavior early warning and monitoring module: constructs a virtual behavior prediction chain to monitor the actual continuous behavior data of the target individual in real time; presets thresholds, analyzes and performs intelligent identification and early warning.

8. The intelligent identification system for smart camp defense monitoring based on deep learning according to claim 7, characterized in that: The dynamic scene coordinate construction module includes an occlusion acquisition unit and a coordinate system construction unit; The obstruction acquisition unit: Based on the area of ​​the camp, deploy several fixed monitoring cameras, and use multi-camera collaborative perception technology in the camp defense monitoring system to collect monitoring images of the camp in real time; based on deep learning and image perception technology, identify temporary obstructions in the camp and their location information in the monitoring images, wherein the temporary obstructions include training equipment piles and vehicles. The coordinate system construction unit is based on the location information of all positions within the camp, the location information of temporary cover, and the Standard Behavior Chain (SBC). i,j Construct a two-dimensional scattered coordinate system for the campsite, as follows: Based on the campsite's area, construct a two-dimensional scattered coordinate system for the campsite, and define the Standard Behavior Chain (SBC). i,j movement path MP i,j The movement path is represented as a sequence of discrete coordinate points in the camp's two-dimensional scattered coordinate system; the location information of the temporary obstruction is mapped to the camp's two-dimensional scattered coordinate system to form a coordinate region, and the range of the x-coordinate of the coordinate region is denoted as [x...]. a,min ,x a,max The range of the ordinates of the coordinate region is denoted as [y]. a,min ,y a,max ].

9. The intelligent identification system for smart camp defense monitoring based on deep learning according to claim 8, characterized in that: The path conflict region identification module includes a virtual behavior prediction chain construction unit and an identification unit; The virtual behavior prediction chain construction unit: based on the camp's two-dimensional scattered coordinate system, if x a,min ≤x n ≤x a,max , and y a,min ≤y n ≤y a,max Then determine if the a-th temporary occlusion has a movement path MP. i,j The virtual behavior prediction chain of the target individual from the i-th position to the j-th position is constructed as follows: the set of intersection path coordinate points of the discrete coordinate point sequence of the target individual's movement path from the i-th position to the j-th position and the coordinate region of the temporary obstruction is constructed. The calibration unit: A preset extension distance r is used to add an extension distance r to the coordinate region of the temporary obstruction. The extended coordinate region is recorded as a detour candidate coordinate region. A set of key coordinate anchor points is calibrated on the boundary of the detour candidate coordinate region, specifically as follows: Based on the horizontal and vertical coordinate ranges of the detour candidate coordinate region; the first and last coordinate points intersecting the discrete coordinate point sequence of the target individual's movement path from the i-th position to the j-th position with the detour candidate coordinate region are extracted and recorded as the initial intersection coordinate points (x, y, y). in ,y in ) and the final intersection point (x out ,y out ).

10. The intelligent identification system for smart camp defense monitoring based on deep learning according to claim 9, characterized in that: The detour route planning and correction module includes a detour route planning unit and a correction unit; The detour path planning unit: selects from the key coordinate anchor point set KSA the coordinate points (x) that intersect with the starting point. in ,y in ) and the final intersection point (x out ,y out The nearest critical coordinate anchor point is denoted as the starting detour coordinate point KSA. in and the final detour coordinates KSA out ; Starting from the coordinate point KSA in Starting from point KSA, the coordinates of the detour end point KSA. out Using the destination as the endpoint, and combining the camp's two-dimensional scattered coordinate system and the candidate detour coordinate region, the A* algorithm is used to generate candidate detour path segments, denoted as R. in-out The discrete coordinate sequence ZMP of the movement path of the target individual from the i-th position to the j-th position. i,j The initial intersection point (x) in ,y in ) to the final intersection point (x out ,y out The coordinate point sequence of ) is replaced with candidate detour path segment R. in-out And marked as the virtual modified movement path (DMP) of the target individual from the i-th position to the j-th position. i,j ; The correction unit is based on standard behavior chains and virtual corrected movement paths (DMP). i,j Calculate the virtual corrected movement time of the target individual from the i-th position to the j-th position.

Images