Multi-agent fusion perception regional intrusion anomaly alarm method and system

Abstract

The application relates to the technical field of motion detection alarm, in particular to a regional intrusion abnormality alarm method and system based on multi-agent fusion perception, which comprises the following steps: collecting target coordinates and time stamps, integrating trajectories and sorting, analyzing direction and time change to determine stability, identifying induced behaviors, screening and verifying abnormalities, and generating alarm instruction output results; in the application, the spatial positions and time information of multi-source perception agents are integrated, the continuous trajectory of a target is reconstructed, the integrity identification of motion behaviors is enhanced, the perception depth of behavior trends is improved through linkage analysis of trajectory direction change and time interval, multi-dimensional identification of abnormal behaviors is realized through linkage judgment of trajectory stability and regional stay state, the persistence condition is introduced in abnormality verification to filter incidental interference, the alarm output is matched based on the verification result and the regional strategy, the accuracy and robustness of intrusion detection are effectively improved, and the practical requirements in a high dynamic scene are met.

Description

Technical Field

[0001] This invention relates to the field of motion detection alarm technology, and in particular to a method and system for alarming regional intrusion anomalies using multi-agent fusion perception. Background Technology

[0002] The field of motion detection alarm technology involves the system design and methodology research for real-time perception and discrimination of target motion behavior within a specific area, and triggering alarm signals accordingly. Core aspects of this technology include target motion feature recognition, multi-source information acquisition and processing, spatial location analysis, and alarm response mechanism construction. Typically, it involves deploying detection devices such as infrared sensors, ultrasonic sensors, and video surveillance equipment to acquire moving target data and analyzing and determining the presence of intrusion behavior based on preset thresholds or behavioral patterns. This type of technology is widely used in security monitoring, boundary protection, and intelligent inspection scenarios. Traditional multi-agent fusion perception-based area intrusion anomaly alarm methods refer to alarm methods that use multiple sensing agents to collaboratively perceive and judge the behavior of targets within an area. The technical challenge is to improve the accuracy and reliability of area intrusion detection in dynamic environments. Traditional multi-agent fusion perception typically employs distributed deployment of multiple infrared or ultrasonic sensors and uses a detection mechanism based on prior rules to monitor targets and determine intrusions. This process generally uses static perception information fusion and target state determination based on simple probability models to achieve information sharing and collaborative judgment among multiple agents, thereby identifying and alarming intrusion behavior.

[0003] Traditional multi-agent fusion perception methods often employ distributed infrared or ultrasonic sensors to acquire static trigger information and make judgments based on preset rules. This lack of ability to reconstruct and analyze the continuous behavioral trajectories of targets easily leads to fragmented behavior and misidentification. In actual operation, due to the discontinuity of sensor data and the influence of occlusion, target states often experience temporary interruptions. The system relies solely on single-point information for judgment, making it difficult to accurately respond to complex behavioral changes. In-depth analysis of trajectory direction changes and area dwelling behavior is not conducted, ignoring the induction paths and regional intent expressions of abnormal behavior, resulting in insufficient ability to identify latent intrusions. Information fusion often uses static probability models, failing to combine real-time behavioral trend evolution for comprehensive judgment, and lacking effective assessment of behavioral persistence and abnormal dwelling. For example, when frequent loitering behavior exists in boundary areas, the system cannot identify its abnormal characteristics through time-based behavioral chains, easily misjudging it as normal movement, or conversely, frequently triggering false alarms and interfering with normal operation. Because alarm triggering strategies are highly dependent on set rules and lack empirical support for behavior, the system cannot form effective adaptive strategies for target behavior in changing scenarios, which reduces the overall security response efficiency and scenario coverage. Summary of the Invention

[0004] To address the technical problems existing in the prior art, embodiments of the present invention provide a method for regional intrusion anomaly alarm based on multi-agent fusion perception, comprising the following steps:

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a multi-agent fusion perception method for regional intrusion anomaly alarm, comprising the following steps:

[0006] S1: Obtain the spatial coordinates and timestamps of the same target collected by multiple sensing agents within the protected area, integrate the trajectory points according to the target identifier and sort them by time, process the displacement and time relationship between trajectory points, and generate a motion trajectory record set;

[0007] S2: Select continuous trajectory segments from the motion trajectory record set, process the changes in motion direction between trajectory segments, generate a trajectory change description in combination with the time interval, compare the trajectory change with the normal direction change threshold, determine the trajectory stability state, and generate a stability judgment result.

[0008] S3: Based on the stability determination results, filter abnormal targets, extract the defense zone division markers from the motion trajectory records, sort out the entry and stay status of targets in the defense zone, combine the entry distribution and stay changes, determine the regional behavior inducement characteristics, and generate inducement behavior recognition results;

[0009] S4: Combine the induced behavior recognition results with trajectory stability to screen abnormal condition targets and record the state duration. When the abnormal duration meets the dwell time threshold, complete the abnormal verification and generate the intrusion abnormal verification result.

[0010] S5: Based on the intrusion anomaly verification result, obtain the alarm zone identifier where the target is located, match the verification result with the zone alarm triggering condition, generate an alarm command identifier and output it to the intrusion anomaly alarm interface to form a regional intrusion anomaly alarm result.

[0011] As a further aspect of the present invention, the motion trajectory record set includes target identifier, trajectory point sequence, timestamp sequence, displacement state parameters, and continuous trajectory index; the stability determination result includes trajectory segment identifier, direction change amount, time interval value, threshold comparison conclusion, and stable state label; the induced behavior identification result includes abnormal target identifier, defense zone division identifier, entry distribution characteristics, dwelling state characteristics, and induced feature label; the intrusion anomaly verification result includes anomaly candidate identifier, anomaly duration, dwell time threshold matching conclusion, anomaly verification status mark, and verification record identifier; and the area intrusion anomaly alarm result includes alarm defense zone identifier, alarm command identifier, alarm trigger matching result, alarm interface output status, and alarm event identifier.

[0012] As a further aspect of the present invention, the specific steps of S1 are as follows:

[0013] S101: Obtain the target location coordinates and timestamps collected by multiple sensing agents within the protected area, classify the data according to the target identifier, organize and filter coordinate points with different sources but the same identifier in chronological order, and generate a joint target coordinate sequence;

[0014] S102: Based on the target joint coordinate sequence, multiple sets of coordinates at the same time point are fused, continuous fused position points are extracted and arranged in chronological order to generate a time series fused trajectory set;

[0015] S103: Based on the spatial position changes and time intervals between adjacent points in the time series fusion trajectory set, calculate the corresponding instantaneous velocity sequence, identify the displacement state by combining the velocity change trend, and generate a motion trajectory record set.

[0016] As a further aspect of the present invention, the specific steps of S2 are as follows:

[0017] S201: Select a continuous trajectory segment based on the motion trajectory record set, call the coordinate parameters between the starting position point and the ending position point within the trajectory segment, extract the motion direction value of the continuous trajectory segment, arrange the motion direction value in time order and integrate it to generate a continuous trajectory direction sequence.

[0018] S202: Based on the continuous trajectory direction sequence, calculate the direction change angle between adjacent trajectory segments, and combine the time interval value between adjacent trajectory segments to process the correspondence between the direction change angle and the time interval to generate a direction change rate sequence.

[0019] S203: Based on the comparison between the values ​​in the direction change rate sequence and the normal walking direction change threshold, determine whether the direction change in the interval of the continuous trajectory segment is within the threshold limit range, obtain the state identifier set of the trajectory segment, and generate the stability determination result.

[0020] As a further aspect of the present invention, the specific steps of S3 are as follows:

[0021] S301: Based on the stability determination result, filter the target number that is determined to be in an abnormal state, call the motion trajectory record set of the corresponding target, extract the defense zone division identifier parameters included in the trajectory node, organize the corresponding records of different defense zone identifiers and target numbers according to the time order, and generate a target defense zone association sequence.

[0022] S302: Based on the target zone association sequence, extract the entry time and exit time of the target number in the zone, calculate the dwell time value of the target in the corresponding zone, and classify the target number, zone number and dwell time value to generate a zone dwell time distribution value.

[0023] S303: Based on the target's entry order and dwell time value in the adjacent defense zones in the defense zone dwell time distribution value, determine whether an abnormal entry pattern is formed in the specified defense zone, and count the consecutive number of abnormal defense zone entry segments. Based on the statistical results, establish a set of behavior change states and generate induced behavior recognition results.

[0024] As a further aspect of the present invention, the specific steps of S4 are as follows:

[0025] S401: Extract the target number based on the induced behavior recognition result and the stability determination result, call the state identifier parameter in the two results, filter the records under the corresponding target number whose state is abnormal, organize the abnormal target numbers in chronological order, and generate a joint abnormal target list;

[0026] S402: Extract the start time and end time of each target number in the abnormal state according to the joint abnormal target list, calculate the duration value of the corresponding abnormal state, and collect the duration values ​​corresponding to the time period according to the target number to generate an abnormal state duration sequence.

[0027] S403: Based on the comparison between each duration value in the abnormal state duration sequence and the set dwell time threshold, mark all target numbers greater than the dwell time threshold as valid abnormal record numbers, organize the corresponding abnormal behavior time period information according to the number, and generate intrusion abnormal verification results.

[0028] As a further aspect of the present invention, the specific steps of S5 are as follows:

[0029] S501: Extract the abnormal target number based on the intrusion anomaly verification result, call the trajectory node data corresponding to the target number, filter the zone division identifier parameters attached to the node, summarize the correspondence between the target number and the zone identifier, and generate alarm zone mapping value.

[0030] S502: Based on the alarm zone mapping value, extract the trajectory dwell time of the target number in the zone, the zone entry frequency and the trigger node density, match and judge against the alarm trigger condition value configured in the zone, filter the target numbers that meet the conditions, and generate alarm trigger matching results;

[0031] S503: Extract the target number and the identification of the zone according to the alarm trigger matching result, construct the alarm command identification and output it to the area alarm command interface, and record the interface response time and alarm level identification to generate the area intrusion anomaly alarm result.

[0032] As a further aspect of the present invention, the plurality of sensing agents refer to sensing nodes deployed within the protected area that are capable of independently collecting the spatial position coordinates of moving targets and generating corresponding timestamps.

[0033] The protected area refers to the monitoring space range pre-set by the regional intrusion anomaly alarm system, and the range is defined by the regional boundary division or the protection zone identification;

[0034] The spatial position coordinates refer to the coordinate data used to represent the position of a moving target within the protected area, which is generated based on a unified coordinate system;

[0035] The timestamp refers to the time stamp generated when the sensing agent collects the corresponding spatial location coordinates;

[0036] The normal direction change threshold refers to the direction change reference limit preset by the system, which is used to limit the range of direction changes of a moving target in the normal travel state within the protected area.

[0037] As a further aspect of the present invention, the zone division identifier refers to the zone identifier information used to distinguish different monitoring sub-regions within the protected area and associated with the trajectory node;

[0038] The so-called stationary state refers to a state in which a moving target is continuously in a positionally restricted state within the same defense zone;

[0039] The alarm zone identifier refers to the area identifier in the zone division identifier that is associated with the alarm triggering rule.

[0040] A multi-agent fusion perception regional intrusion anomaly alarm system includes:

[0041] The data acquisition module is used to execute S1: acquire the spatial position coordinates and corresponding timestamps of multiple sensing agents in the protected area for the same moving target, integrate the trajectory points according to the target identifier and arrange them in chronological order, process the displacement state and time relationship of continuous trajectory points, form a description of the continuous motion trajectory of the target in the protected area, and generate a motion trajectory record set.

[0042] The trajectory construction module is used to execute S2: select continuous trajectory segments based on the target area motion trajectory record set, process the changes in motion direction between trajectory segments, and form a trajectory change description in combination with the corresponding time interval. The trajectory change is compared with the normal walking direction change threshold to complete the trajectory stability judgment and generate a stability judgment result.

[0043] The stability determination module is used to execute S3: based on the stability determination result, filter targets with abnormal trajectory stability, call the corresponding target area motion trajectory record set to obtain the defense zone division identifier associated with the trajectory node, organize the target's entry and stay status in the defense zone, complete the regional behavior guidance feature judgment based on the changes in the defense zone entry distribution and stay status, and generate the guidance behavior recognition result;

[0044] The behavior recognition module is used to perform S4: based on the induced behavior recognition result and the stability judgment result, perform joint judgment, include targets that meet the abnormal conditions into the abnormal candidate range, and record the duration of the abnormal state. When the duration meets the dwell time threshold requirement, complete the abnormal verification and generate the intrusion abnormal verification result.

[0045] The alarm generation module is used to execute S5: obtain the alarm zone identifier of the target based on the intrusion anomaly verification result, match the anomaly verification result with the zone alarm triggering condition, form a corresponding alarm instruction identifier and output it to the area intrusion anomaly alarm interface to generate the area intrusion anomaly alarm result.

[0046] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0047] In this invention, by integrating the spatial location and temporal information of multi-source sensing agents, the continuous trajectory of the target is reconstructed, enhancing the integrity recognition of motion behavior. The linkage analysis of trajectory direction changes and time intervals improves the depth of perception of behavioral trends. The linkage judgment of trajectory stability and regional dwell state realizes multi-dimensional recognition of abnormal behavior. The anomaly verification introduces continuous conditions to filter occasional interference. The alarm output is based on the verification results and regional strategy matching, effectively improving the accuracy and robustness of intrusion detection and meeting the practical needs of high dynamic scenarios. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 This is a schematic diagram of the steps of the present invention;

[0050] Figure 2 This is a detailed schematic diagram of S1 of the present invention;

[0051] Figure 3 This is a detailed schematic diagram of S2 of the present invention;

[0052] Figure 4This is a detailed schematic diagram of S3 of the present invention;

[0053] Figure 5 This is a detailed schematic diagram of S4 of the present invention;

[0054] Figure 6 This is a detailed schematic diagram of S5 of the present invention. Detailed Implementation

[0055] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0056] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0057] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0058] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0059] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0060] Please see Figure 1 This invention provides a method for alarming regional intrusion anomalies through multi-agent fusion perception, comprising the following steps:

[0061] S1: Obtain the spatial position coordinates and corresponding timestamps of multiple sensing agents within the protected area for the same moving target, integrate the trajectory points according to the target identifier and arrange them in chronological order, process the displacement state and time relationship of continuous trajectory points, form a description of the continuous motion trajectory of the target within the protected area, and generate a motion trajectory record set.

[0062] Multiple sensing agents refer to sensing nodes deployed within the protected area that can independently collect the spatial coordinates of moving targets and generate corresponding timestamps. The collection results are used to describe the displacement of moving targets within the protected area.

[0063] A protected area refers to the monitoring space range pre-defined by the regional intrusion anomaly alarm system. The range is defined by the area boundary or the zone marking, and is used to determine whether a moving target is within the intrusion monitoring range.

[0064] A moving target refers to a monitored object that generates continuous displacement within the protected area and whose spatial location information is collected by a sensing agent. The displacement information is used to construct the target's motion trajectory.

[0065] Spatial position coordinates refer to coordinate data used to represent the position of a moving target within a protected area. They are generated based on a unified coordinate system and are used to describe the changes in the spatial position of a moving target at different points in time.

[0066] A timestamp is a time stamp generated when a sensing agent collects the corresponding spatial location coordinates, used to determine the temporal order between trajectory nodes;

[0067] S2: Select continuous trajectory segments based on the target area motion trajectory record set, process the changes in motion direction between trajectory segments, and form a trajectory change description in combination with the corresponding time interval. Compare the trajectory change with the normal walking direction change threshold to complete the trajectory stability judgment and generate stability judgment result.

[0068] The normal walking direction change threshold refers to the direction change reference limit preset by the system, which is used to limit the range of direction changes of a moving target in the normal walking state within the protected area, and serves as a reference for judging the trajectory stability state.

[0069] S3: Based on the stability judgment results, filter targets with abnormal trajectory stability, call the corresponding target area motion trajectory record set to obtain the defense zone division identifier associated with the trajectory node, sort out the target's entry and stay status in the defense zone, complete the regional behavior guidance feature judgment based on the changes in the defense zone entry distribution and stay status, and generate the guidance behavior recognition result;

[0070] Zone division markers refer to the zone identification information used to distinguish different monitoring sub-areas within a protected area and are associated with trajectory nodes to determine the zone location of a moving target within the protected area;

[0071] The loitering state refers to the state in which a moving target is continuously in a restricted position within the same defense zone, and is used to describe the loitering behavior of a moving target within the defense zone;

[0072] S4: Based on the results of induced behavior recognition and stability determination, a joint determination is made, targets that meet the abnormal conditions are included in the abnormal candidate range, and the duration of the abnormal state is recorded. When the duration meets the dwell time threshold requirement, the abnormal verification is completed and the intrusion abnormal verification result is generated.

[0073] The dwell time threshold refers to the time reference limit set in advance by the system, which is used to determine whether the duration of a moving target's stay in the defense zone constitutes abnormal dwell behavior.

[0074] S5: Obtain the alarm zone identifier of the target based on the intrusion anomaly verification result, match the anomaly verification result with the zone alarm trigger condition, form the corresponding alarm command identifier and output it to the area intrusion anomaly alarm interface to generate the area intrusion anomaly alarm result.

[0075] Alarm zone identifiers refer to the area identifiers in the zone division identifiers that are associated with the alarm triggering rules, and are used to determine the zone range corresponding to the intrusion anomaly alarm output.

[0076] The motion trajectory record set includes target identifier, trajectory point sequence, timestamp sequence, displacement state parameters, and continuous trajectory index. The stability judgment result includes trajectory segment identifier, direction change, time interval value, threshold comparison conclusion, and stable state label. The induced behavior recognition result includes abnormal target identifier, defense zone division identifier, entry distribution characteristics, dwelling state characteristics, and induced feature label. The intrusion anomaly verification result includes anomaly candidate identifier, anomaly duration, dwell time threshold matching conclusion, anomaly verification status mark, and verification record identifier. The area intrusion anomaly alarm result includes alarm defense zone identifier, alarm command identifier, alarm trigger matching result, alarm interface output status, and alarm event identifier.

[0077] Please see Figure 2 The specific steps of S1 are as follows:

[0078] S101: Obtain the target location coordinates and timestamps collected by multiple sensing agents within the protected area, classify the data according to the target identifier, organize and filter coordinate points with different sources but the same identifier in chronological order, and generate a joint target coordinate sequence;

[0079] Multiple intelligent agents with sensing capabilities are deployed within the protected area. These agents include millimeter-wave radar, infrared detection equipment, optical camera systems, and sonar sensing nodes. Each agent records the collected target spatial coordinates and time information in real time through a pre-set unified time synchronization mechanism, such as using a high-precision time synchronization protocol. Each record contains a unique target identifier, three-dimensional spatial coordinates, and corresponding time information. All data is uploaded to an information fusion processing platform. The system categorizes the received raw sensing data according to the target identifier, using the target ID as a keyword to construct a data structure and store information about the same target from different sources. Data belonging to the same identifier in each category is sorted according to timestamps. The sorting algorithm arranges the data in ascending order of time. Further validity screening is performed on the categorized data, and spatial consistency checks are conducted on data points with similar times. With a time error tolerance of 0.1 seconds, coordinate spatial differences are detected for the same target data collected by different sensing agents, provided the time difference meets the requirements. If the spatial position deviation exceeds the set tolerance standard of 0.5m, the data set is considered an error value and discarded, or marked as low-confidence data. During the screening process, the three-dimensional coordinates of each pair of data are analyzed for their acceptability. The discarded data are stored uniformly according to the target identifier, ultimately forming a joint position sequence for each target at continuous time points with high consistency of origin and accurate spatial matching, which is used for subsequent trajectory fusion and dynamic analysis.

[0080] S102: Based on the target joint coordinate sequence, multiple sets of coordinates at the same time point are fused, continuous fused position points are extracted and arranged in chronological order to generate a time series fused trajectory set;

[0081] Based on the constructed target joint coordinate sequence, data from multiple sensing sources need to be fused at the same time point. The confidence level of each sensing node's values ​​differs, requiring confidence weights to be set according to device type and actual operating status. For example, millimeter-wave radar is weighted at 40%, infrared sensing at 30%, optical cameras at 20%, and other types at 10%, totaling 100%. At the same time, the spatial position data collected by each sensor are weighted according to the set weights to obtain fused coordinate values ​​in the horizontal, vertical, and longitudinal directions. All fused coordinate points are connected sequentially in time to form a continuous three-dimensional spatial trajectory. For example, at 13 seconds, three sensors collect a set of horizontal values ​​with coordinates of 13.2, 13.3, and 13.1 (in meters), corresponding to weights of 40%, 30%, and 30%. The fused horizontal coordinates can then be weighted proportionally to obtain an equilibrium value. Similarly, the longitudinal and vertical information are processed to complete the overall fused point construction. After the fused trajectory is constructed, the update interval between each time point needs to be detected to ensure the continuity of the trajectory sequence. For example, a maximum interval of 0.5 seconds can be set. If the time interval between two points exceeds this threshold, it is considered a trajectory interruption point and needs to be supplemented or marked as interrupted. The final set of time-series fused trajectories is uniformly recorded as a spatial point sequence with time labels, which can be directly used for dynamic behavior determination and movement feature analysis.

[0082] S103: Based on the spatial position changes and time intervals between adjacent points in the time series fusion trajectory set, calculate the corresponding instantaneous velocity sequence, identify the displacement state by combining the velocity change trend, and generate a motion trajectory record set;

[0083] Motion characteristic calculations are performed on continuous points in the fused trajectory, requiring the calculation of the target's instantaneous velocity based on the spatial and temporal changes between adjacent points. For any two consecutive fused points, the spatial coordinate difference is calculated, and then compared with its time interval to obtain the corresponding velocity value. For example, if one coordinate point has a time interval of 13.0 seconds and another has a time interval of 13.2 seconds, with a spatial position difference of 0.6m, then its corresponding velocity is 3m per second. This calculation operation is repeated throughout the entire trajectory sequence to obtain a complete list of velocity changes. Further, the movement state needs to be determined based on the velocity change trend. The movement state is divided into stationary, uniform movement, and accelerated movement. The threshold for stationary movement is set at 0.2m / s, the uniform movement range is 0.2 to 1.5m / s, and acceleration is defined as exceeding 1.5m / s. Each velocity value is classified according to a set standard. A sliding window method is used for the velocity sequence, with a window length of 5. The average velocity within the window is calculated, and the difference is analyzed with the previous window. If the average velocity increases by more than 0.5m / s, it can be judged as an acceleration phase; if it decreases by more than the same amount, it is a deceleration phase. The velocity value and the state recognition result are recorded together in the trajectory state table, forming a comprehensive record sequence containing position, time, velocity and state information, providing input data support for dynamic behavior recognition and subsequent tracking decisions.

[0084] Please see Figure 3 The specific steps of S2 are as follows:

[0085] S201: Select a continuous trajectory segment based on the motion trajectory record set, call the coordinate parameters between the starting and ending points within the trajectory segment, extract the motion direction value of the continuous trajectory segment, arrange the motion direction value in chronological order and integrate it to generate a continuous trajectory direction sequence.

[0086] Continuous trajectory segments are selected based on the motion trajectory record set. Each trajectory segment must consist of two or more trajectory points with a temporal relationship. The selection criteria are that each segment contains a time span of no less than 2 seconds and at least 3 sets of trajectory points. The three-dimensional coordinate values ​​of the starting point and the ending point are extracted and denoted as the starting point coordinates (x1, y1) and the ending point coordinates (x2, y2), respectively. A straight line direction vector is constructed based on these two coordinate points, and the motion direction value of the trajectory segment is extracted. This direction value is the angle between the line connecting the starting and ending points and the due north direction. The direction value is calculated using the arctangent function, i.e., the direction angle is the angle between the vector and the y-axis. The angle, combined with the coordinate difference, can be simplified according to the trajectory direction formula. For example, if the starting point of a trajectory is (12, 15) and the ending point is (16, 18), then the lateral displacement is 4 and the longitudinal displacement is 3. Based on this difference, the direction value is calculated to be approximately 53.1 degrees. The direction value of each trajectory segment is time-stamped and sorted by time. During the sorting process, an interpolation method is used to correct the time of the data points. When there are different trajectory segments with direction values ​​within the same time period, the rule of prioritizing the retention of the largest timestamp is adopted. Finally, the direction values ​​of all continuous trajectory segments are integrated to generate a continuous trajectory direction sequence containing the time series.

[0087] S202: Based on the continuous trajectory direction sequence, calculate the angle of change of direction between adjacent trajectory segments, and combine the time interval between adjacent trajectory segments to process the correspondence between the angle of change of direction and the time interval to generate a sequence of direction change rate.

[0088] Based on the continuous trajectory direction sequence, the angle of directional change between adjacent trajectory segments is calculated. This involves subtracting the values ​​of two adjacent directions in the sequence to obtain the angle of directional change, expressed as an absolute value. For example, if the direction value of one segment is 60 degrees and the next segment is 75 degrees, the angle of directional change is 15 degrees. Then, the time interval between adjacent trajectory segments is extracted, which is the start time of the next segment minus the end time of the current segment. For instance, if the end time of the current segment is 14.0 seconds and the start time of the next segment is 14.3 seconds, the time interval is 0.3 seconds. The angle of directional change and the time interval are logarithmically calculated, and the rate of directional change is obtained by comparing these two values. The rate is measured in degrees per second. Each group of directional change rate values ​​is sorted, and a directional change rate sequence is generated using the timestamp as the key. If the directional angle is found to be greater than 180 degrees during sorting, angle rotation processing is required. For example, if the change is from 170 degrees to 10 degrees, the actual change angle should be processed according to the minimum angle of 30 degrees. Each group of directional change rate values ​​is divided into zones: the low-speed zone is set to 0 to 20 degrees per second, the medium-speed zone is 20 to 60 degrees per second, and the high-speed zone is above 60 degrees per second. The data is recorded in a unified format as a sequence data structure, which includes the start and end time of each segment, the directional difference, the time interval, and the corresponding rate, ultimately forming a directional change rate sequence.

[0089] S203: Based on the comparison between the values ​​in the direction change rate sequence and the normal walking direction change threshold, determine whether the direction change of the interval in the continuous trajectory segment is within the threshold limit range, obtain the state identifier set of the trajectory segment, and generate the stability judgment result.

[0090] The system compares the rate values ​​of each group in the direction change rate sequence with the normal walking direction change threshold. The threshold is set with reference to the statistical results of actual human walking direction changes. For example, in an urban walking environment, the normal walking direction change rate usually does not exceed 40 degrees per second. Therefore, the upper limit of the direction change threshold is set to 40 degrees per second. The judgment process uses each rate value as the judgment object. If the direction change rate of a certain segment is 15 degrees per second, it is judged to be within the normal range. If it is 45 degrees per second, it exceeds the threshold and needs to be marked as an abnormal state. A status label, such as normal or abnormal, is added to each segment result and combined with the trajectory segment number information to form a status label set. A buffer mechanism needs to be set in the judgment process. If only one of the three adjacent trajectory segments is abnormal, it is not considered as an overall abnormality. A minimum number of consecutive abnormal segments needs to be set. For example, it is set that two or more consecutive segments exceed the threshold before it is judged as an unstable state. Based on actual numerical calculations, the sample trajectory segment numbers can be set as 1, 2, and 3, with direction change rates of 35, 48, and 52 degrees per second, respectively. Then, the numbers 2 and 3 exceed the threshold consecutively, and the whole is marked as unstable. Finally, the stability judgment result is generated by arranging the status labels of all trajectory segments in chronological order.

[0091] Please see Figure 4 The specific steps of S3 are as follows:

[0092] S301: Based on the stability judgment result, filter the target number that is judged to be in an abnormal state, call the motion trajectory record set of the corresponding target, extract the defense zone division identifier parameters included in the trajectory node, organize the corresponding records of different defense zone identifiers and target numbers according to the time order, and generate the target defense zone association sequence.

[0093] Based on the stability assessment results, target numbers that are determined to be in an abnormal state are filtered. First, the stable state result table is searched, and target numbers whose status identifier field value is not "normal" are extracted and summarized. For example, in a certain stable state dataset, targets numbered A001, A003, and A006 are marked as abnormal. These numbers are then used as subsequent extraction objects. When calling the motion trajectory record set of the corresponding target, the corresponding complete trajectory data structure needs to be located according to the target number. This structure should contain basic fields such as time, spatial coordinates, and zone number. The zone division identifier parameter associated with each trajectory point is extracted from it. For example, the zone number of target A001 at 13:00:02 is Z01. At 13:00:06, the target enters Z02 and continues until 13:00:12, then transitions to Z03. After all trajectory nodes are sorted in ascending order of time, the combination information of their corresponding zone number and target number is extracted one by one to form a sequence of zone changes of the target in the time dimension. If there is a time continuity interruption or a jump in zone number in this sequence, it needs to be supplemented. For example, if the target trajectory jumps from Z01 to Z03 and lacks the intermediate segment Z02, the zone identifier is supplemented by cross-judgment based on the coordinates of the trajectory points and the coordinates of the zone boundaries. Finally, the zone numbers of all trajectory nodes are grouped and organized according to the target number to generate a time sequence mapping data structure between different zones in the complete movement path of each abnormal target, thus obtaining the target zone association sequence.

[0094] S302: Based on the target zone association sequence, extract the entry time and exit time of the target number in the zone, calculate the dwell time value of the target in the corresponding zone, and classify the target number, zone number and dwell time value to generate the zone dwell time distribution value.

[0095] Based on the target zone association sequence, for each target number, the entry and exit time points within each zone are extracted according to the zone number. The entry time is the time when the trajectory of the first appearance of the zone identifier, and the exit time is the time of the last appearance of the zone identifier plus the minimum sampling interval value set by the system. If the sampling interval is 0.2 seconds, then the last trajectory time of target A001 in zone Z01 is 13:00:10, and the exit time is 13:00:10 plus 0.2 seconds, i.e., 13:00:10.2. The dwell time value is calculated by subtracting the entry time from the exit time. For example, if the entry time is... If the exit time is 13:00:02 and the exit time is 13:00:10.2, then the dwell time is 8.2 seconds. The target number, zone number, and corresponding dwell time are recorded together. At the same time, when classifying, multiple entries of each target in the same zone are merged and the sum of the dwell times of each segment is taken as the final dwell time value. If target A003 has two dwell times of 5 seconds and 3 seconds in zone Z02, then the merged record is 8 seconds. The dwell time results of all target numbers are classified according to the zone number. Multiple targets and their corresponding dwell times are recorded under each category to form the zone dwell time distribution value.

[0096] S303: Based on the target's entry order and dwell time value in adjacent defense zones in the zone dwell time distribution value, determine whether an abnormal entry pattern is formed in the specified defense zone, and count the consecutive number of abnormal zone entry segments. Based on the statistical results, establish a set of behavior change states and generate induced behavior recognition results.

[0097] Based on the target's entry order and dwell time value in adjacent defense zones according to the zone dwell time distribution value, the defense zone sequence of each target number is first sorted by time. The transition relationship between adjacent defense zone numbers is then determined. For example, if target A005 dwells for 5 seconds in Z01, then enters Z03 for 1 second, and then enters Z02 for 9 seconds, its zone entry order is Z01→Z03→Z02. Combined with the set abnormal entry mode judgment criteria, the standard is that switching between non-contiguous numbered defense zones with a dwell time of less than 3 seconds is considered an abnormal mode. For example, consecutive numbering is defined as Z01 to Z02 and Z02 to Z03 as normal jumps. If there is a jump from Z01 to Z03 or from Z03 to Z01 without any expected dwell time in between, it is considered an abnormal mode. Entry is considered normal, and a stay of less than 3 seconds is considered short-term probing behavior. Meeting both conditions is recorded as an abnormal entry segment. The number of abnormal segments is counted. If there are more than 3 abnormal entry segments for the same target, it is marked as "high-frequency abnormal switching" in the behavior state set. If there are 1 to 2 segments, it is marked as "suspicious entry". Otherwise, it is "stable mode". For example, if target A007 has 4 abnormal switching segments in its trajectory, Z01→Z03, Z03→Z01, Z02→Z04, and Z04→Z02, the count is 4 and the status is marked as "high-frequency abnormal switching". The behavior state of each target number is associated with its corresponding trajectory segment number and recorded in the behavior change state set to finally generate the induced behavior recognition result.

[0098] Please see Figure 5 The specific steps of S4 are as follows:

[0099] S401: Extract the target number based on the induced behavior recognition result and the stability judgment result, call the status identifier parameter in the two results, filter the records under the corresponding target number whose status is abnormal, sort the abnormal target numbers in chronological order, and generate a joint abnormal target list;

[0100] Based on the results of induced behavior recognition and stability determination, target numbers are extracted. First, two independent state datasets are established, named the Induced Behavior State Table and the Stable State Table, respectively. Each table contains target number, state identifier, and state time period fields. The set of target numbers with the state field "abnormal" is selected from the Induced Behavior State Table, denoted as set A. Then, the set of target numbers with the state field "abnormal" is extracted from the Stable State Table, denoted as set B. An intersection operation is performed on sets A and B; the target numbers that exist in both sets are taken as the selection result. For example, set A contains A001, A002, and A005, and set B contains A002, A003, A004, A005, A006, A007, A008, A009 ... 5. If A007 is the target, then the intersection is A002 and A005. Next, the abnormal time period information corresponding to each target number in the intersection is sorted. The sorting basis is the start time of the abnormal state. Assuming that A002 has three abnormal state times, namely 10:01–10:10, 10:15–10:20, and 10:25–10:35, they are sorted in ascending order of start time. Finally, the target numbers and their sorted abnormal time period lists are organized into a structured data table. The record format includes the target number, the start and end time of the abnormal time period, and the state source identifier. The state information from multiple sources is aggregated into one through the time clue to generate a joint abnormal target list.

[0101] S402: Extract the start and end times of each target number under the abnormal state from the joint abnormal target list, calculate the duration value of the corresponding abnormal state, and collect the duration values ​​corresponding to the time period according to the target number to generate an abnormal state duration sequence.

[0102] Based on the joint list of abnormal targets, extract the time periods for each target number when it is marked as abnormal. Obtain the start and end times of each segment, and calculate the duration by subtracting the start time from the end time. The duration is recorded in seconds. For example, if target A002 enters an abnormal state at 10:01:00 and exits at 10:10:00, the duration is 600 seconds. For multiple abnormal state segments for the same target number, record the duration of each segment and summarize them into a list of time periods corresponding to that target number, generating a list such as: A002 → [600, 300, 600] seconds. Using the target number as the primary key, the duration values ​​of the time periods are organized into an array to form an abnormal state duration sequence. During the organization process, overlapping time periods need to be handled. For example, if a segment starts at 10:05 and the previous segment ends at 10:10, there is partial overlap. The two segments should be merged into 10:01–10:20, with a total duration of 1140 seconds, to avoid double-counting the duration. For overlap judgment, a tolerance value of 0.1 seconds is set. If the difference between time periods is less than this value, it is determined to be a continuous state segment, and merging is performed. Finally, the time periods and corresponding duration combinations are output with the target number as the unit to construct a complete abnormal state duration sequence.

[0103] S403: Compare each duration value in the abnormal state duration sequence with the set dwell time threshold, mark all target numbers that are greater than the dwell time threshold as valid abnormal record numbers, organize the corresponding abnormal behavior time period information according to the number, and generate intrusion abnormal verification results;

[0104] The system compares each duration value in the abnormal state duration sequence with a set dwell time threshold of 300 seconds. This threshold is based on the shortest acceptable abnormal dwell time standard in actual scenarios. Each duration value is compared, and if the duration is greater than 300 seconds, it is considered a valid abnormal state. For example, target A002 has three durations of 600, 300, and 600 seconds. Two of these exceed the threshold, and one is equal to the threshold. The rule is set so that anything greater than or equal to the threshold is considered a valid abnormality. Therefore, all three durations of A002 are valid records, recorded as A002→[(10:01–10:10), (10:15–10:20), (10:25–10:35)]. The same judgment is performed on all target numbers, and segments that meet the conditions are selected. The process involves further mapping the start and end times of each abnormal state to the target number, compiling this information into an intrusion anomaly record table. This table includes a field indicating whether the anomaly is a series of consecutive abnormal states. If a single target number experiences three or more consecutive abnormal time periods within 30 minutes, with a cumulative duration exceeding 900 seconds, then that target number is marked as a "high-risk anomaly." For example, target A005 lasts for 300, 240, and 360 seconds respectively at 09:00–09:05, 09:08–09:12, and 09:14–09:20, with each of the three intervals less than 5 minutes and a total duration of 900 seconds. This meets the criteria, so A005 is marked as a high-risk anomaly. Finally, an intrusion anomaly verification result list is generated, including the target number, abnormal segment time, duration, valid identifier, and risk level label.

[0105] Please see Figure 6 The specific steps of S5 are as follows:

[0106] S501: Extract the abnormal target number based on the intrusion anomaly verification result, call the trajectory node data corresponding to the target number, filter the zone division identifier parameters attached to the node, summarize the correspondence between the target number and the zone identifier, and generate alarm zone mapping value;

[0107] Based on the intrusion anomaly verification results, the abnormal target IDs are extracted. First, an intrusion anomaly verification result data table is created, containing fields such as target ID, anomaly start and end time, zone ID, and anomaly judgment identifier. All target IDs identified as abnormal are extracted and deduplicated to generate a unique set of IDs, for example, A101, A105, and A110. Then, the original trajectory node data table corresponding to these IDs is called. All trajectory records are extracted from this table by matching the target ID. Each record should contain a timestamp, spatial coordinates, and zone division identifier parameters. After extraction, the trajectory nodes are sorted in ascending order by time. The zone identifier field in each node record is read and summarized, and the target ID is matched with the target's location in time. The zone numbers appearing in the entire trajectory are paired, and the frequency of the target's appearance in each zone is counted to generate a mapping table between the target and the zone. The format is that the target number corresponds to a set of zone numbers, for example, A101→[Z01, Z02, Z04], A105→[Z03, Z04]. During the processing, the zone numbers are made unique. If the same target appears multiple times in a certain zone, only one number is retained for structural simplification. At the same time, a field is added to the mapping data to indicate whether the target number has cross-zone movement behavior. If there are more than two types of zone numbers, it is marked as "cross-zone", otherwise it is "single-zone". Finally, an alarm zone mapping value containing the target number, the set of zone numbers, and the trajectory time period is formed.

[0108] S502: Based on the alarm zone mapping value, extract the trajectory dwell time of the target number in the zone, the zone entry frequency and the trigger node density, match and judge against the alarm trigger condition value configured in the zone, filter the target numbers that meet the conditions, and generate alarm trigger matching results;

[0109] Based on the alarm zone mapping value, three key parameters are extracted for each target number within its corresponding zone: trajectory dwell time, zone entry frequency, and trigger node density. First, the dwell time within the zone is calculated by extracting the first entry time and last exit time of the same target within a given zone number, and then calculating the time span corresponding to all trajectory points. If target A101 stays in zone Z01 from 12:00:10 to 12:05:30 for a total of 320 seconds, then its dwell time in Z01 is 320 seconds. The entry frequency is calculated as the number of times the target enters the same zone in different time periods. If target A105 has 3 entry records in Z04 with an intermediate exit process, then its entry frequency is 3 times, and the trigger node density is... Point density is calculated by dividing the total number of trajectory points of the target within the zone by the dwell time. For example, if there are 100 trajectory points and the dwell time is 500 seconds, the node density is 0.2 points / s. After calculation, the three indicators are matched with the alarm triggering condition values ​​set for the zone. The triggering condition values ​​need to be set according to the zone level. For example, Z01 sets the dwell time threshold to 300 seconds, the entry frequency to 2 times, and the node density to 0.1 points / s. If the target's three indicators are all higher than the threshold in Z01, it is considered to meet the alarm triggering conditions. All target numbers that meet all three indicators or any two of them are filtered out, and an alarm triggering matching result list is generated. The record format is target number, zone number, value of each indicator, and whether the alarm is triggered.

[0110] S503: Extract the target number and the identification of the zone based on the alarm trigger matching result, construct the alarm command identification and output it to the area alarm command interface, while recording the interface response time and alarm level identification, and generating the area intrusion anomaly alarm result;

[0111] Based on the alarm trigger matching results, the target number and its corresponding zone identifier are extracted. An alarm command identifier is constructed for each set of data. This identifier includes fields such as alarm number, target number, zone number, alarm time, trigger source information, and alarm level. The alarm number can be constructed using a format such as "ALM_Y ... If two conditions are met, it is set to Level 2; if all three conditions are met, it is set to Level 1. After each alarm command is constructed, it is pushed to the regional alarm command interface through the interface standard. At the same time, the interface response time is recorded in the log. For example, if the command is issued at 13:15:31 and the interface returns at 13:15:31.25, the response time is 0.25 seconds. The recorded content includes the interface address, return code, response time, and whether the push was successful. All alarm commands, response information, and alarm levels are generated together into a structured data table, which records the final alarm command issuance status and the spatial area corresponding to the abnormal behavior, forming the regional intrusion anomaly alarm result.

[0112] A multi-agent fusion perception regional intrusion anomaly alarm system includes:

[0113] The data acquisition module is used to execute S1: acquire the spatial position coordinates and corresponding timestamps of multiple sensing agents in the protected area for the same moving target, integrate the trajectory points according to the target identifier and arrange them in chronological order, process the displacement state and time relationship of continuous trajectory points, form a description of the continuous motion trajectory of the target in the protected area, and generate a motion trajectory record set.

[0114] The trajectory construction module is used to execute S2: select continuous trajectory segments based on the target area motion trajectory record set, process the changes in motion direction between trajectory segments, and form a trajectory change description in combination with the corresponding time interval. The trajectory change is compared with the normal walking direction change threshold to complete the trajectory stability judgment and generate a stability judgment result.

[0115] The stability determination module is used to execute S3: based on the stability determination results, it filters targets with abnormal trajectory stability, calls the corresponding target area motion trajectory record set to obtain the defense zone division identifier associated with the trajectory node, sorts out the target's entry and stay status in the defense zone, completes the regional behavior guidance feature judgment based on the changes in the defense zone entry distribution and stay status, and generates guidance behavior recognition results;

[0116] The behavior recognition module is used to execute S4: based on the results of induced behavior recognition and stability judgment, it performs joint judgment, includes targets that meet the abnormal conditions into the abnormal candidate range, and records the duration of the abnormal state. When the duration meets the dwell time threshold requirement, it completes the abnormal verification and generates the intrusion abnormal verification result.

[0117] The alarm generation module is used to execute S5: obtain the alarm zone identifier of the target based on the intrusion anomaly verification result, match the anomaly verification result with the zone alarm trigger condition, form the corresponding alarm instruction identifier and output it to the area intrusion anomaly alarm interface to generate the area intrusion anomaly alarm result.

[0118] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for alarming regional intrusion anomalies using multi-agent fusion perception, characterized in that, Includes the following steps: S1: Obtain the spatial coordinates and timestamps of the same target collected by multiple sensing agents within the protected area, integrate the trajectory points according to the target identifier and sort them by time, process the displacement and time relationship between trajectory points, and generate a motion trajectory record set; S2: Select continuous trajectory segments from the motion trajectory record set, process the changes in motion direction between trajectory segments, generate a trajectory change description in combination with the time interval, compare the trajectory change with the normal direction change threshold, determine the trajectory stability state, and generate a stability judgment result. S3: Based on the stability determination results, filter abnormal targets, extract the defense zone division markers from the motion trajectory records, sort out the entry and stay status of targets in the defense zone, combine the entry distribution and stay changes, determine the regional behavior inducement characteristics, and generate inducement behavior recognition results; S4: Combine the induced behavior recognition results with trajectory stability to screen abnormal condition targets and record the state duration. When the abnormal duration meets the dwell time threshold, complete the abnormal verification and generate the intrusion abnormal verification result. S5: Based on the intrusion anomaly verification result, obtain the alarm zone identifier where the target is located, match the verification result with the zone alarm triggering condition, generate an alarm command identifier and output it to the intrusion anomaly alarm interface to form a regional intrusion anomaly alarm result.

2. The multi-agent fusion perception regional intrusion anomaly alarm method according to claim 1, characterized in that, The motion trajectory record set includes target identifier, trajectory point sequence, timestamp sequence, displacement state parameters, and continuous trajectory index. The stability determination result includes trajectory segment identifier, direction change, time interval value, threshold comparison conclusion, and stable state label. The induced behavior identification result includes abnormal target identifier, defense zone division identifier, entry distribution characteristics, dwelling state characteristics, and induced feature label. The intrusion anomaly verification result includes anomaly candidate identifier, anomaly duration, dwell time threshold matching conclusion, anomaly verification status mark, and verification record identifier. The area intrusion anomaly alarm result includes alarm defense zone identifier, alarm command identifier, alarm trigger matching result, alarm interface output status, and alarm event identifier.

3. The regional intrusion anomaly alarm method according to claim 1, characterized in that, The specific steps of S1 are as follows: S101: Obtain the target location coordinates and timestamps collected by multiple sensing agents within the protected area, classify the data according to the target identifier, organize and filter coordinate points with different sources but the same identifier in chronological order, and generate a joint target coordinate sequence; S102: Based on the target joint coordinate sequence, multiple sets of coordinates at the same time point are fused, continuous fused position points are extracted and arranged in chronological order to generate a time series fused trajectory set; S103: Based on the spatial position changes and time intervals between adjacent points in the time series fusion trajectory set, calculate the corresponding instantaneous velocity sequence, identify the displacement state by combining the velocity change trend, and generate a motion trajectory record set.

4. The regional intrusion anomaly alarm method based on multi-agent fusion perception according to claim 3, characterized in that, The specific steps of S2 are as follows: S201: Select a continuous trajectory segment based on the motion trajectory record set, call the coordinate parameters between the starting position point and the ending position point within the trajectory segment, extract the motion direction value of the continuous trajectory segment, arrange the motion direction value in time order and integrate it to generate a continuous trajectory direction sequence. S202: Based on the continuous trajectory direction sequence, calculate the direction change angle between adjacent trajectory segments, and combine the time interval value between adjacent trajectory segments to process the correspondence between the direction change angle and the time interval to generate a direction change rate sequence. S203: Based on the comparison between the values ​​in the direction change rate sequence and the normal walking direction change threshold, determine whether the direction change in the interval of the continuous trajectory segment is within the threshold limit range, obtain the state identifier set of the trajectory segment, and generate the stability determination result.

5. The regional intrusion anomaly alarm method according to claim 4, characterized in that, The specific steps for S3 are as follows: S301: Based on the stability determination result, filter the target number that is determined to be in an abnormal state, call the motion trajectory record set of the corresponding target, extract the defense zone division identifier parameters included in the trajectory node, organize the corresponding records of different defense zone identifiers and target numbers according to the time order, and generate a target defense zone association sequence. S302: Based on the target zone association sequence, extract the entry time and exit time of the target number in the zone, calculate the dwell time value of the target in the corresponding zone, and classify the target number, zone number and dwell time value to generate a zone dwell time distribution value. S303: Based on the target's entry order and dwell time value in the adjacent defense zones in the defense zone dwell time distribution value, determine whether an abnormal entry pattern is formed in the specified defense zone, and count the consecutive number of abnormal defense zone entry segments. Based on the statistical results, establish a set of behavior change states and generate induced behavior recognition results.

6. The multi-agent fusion perception regional intrusion anomaly alarm method according to claim 5, characterized in that, The specific steps of S4 are as follows: S401: Extract the target number based on the induced behavior recognition result and the stability determination result, call the state identifier parameter in the two results, filter the records under the corresponding target number whose state is abnormal, organize the abnormal target numbers in chronological order, and generate a joint abnormal target list; S402: Extract the start time and end time of each target number in the abnormal state according to the joint abnormal target list, calculate the duration value of the corresponding abnormal state, and collect the duration values ​​corresponding to the time period according to the target number to generate an abnormal state duration sequence. S403: Based on the comparison between each duration value in the abnormal state duration sequence and the set dwell time threshold, mark all target numbers greater than the dwell time threshold as valid abnormal record numbers, organize the corresponding abnormal behavior time period information according to the number, and generate intrusion abnormal verification results.

7. The multi-agent fusion perception regional intrusion anomaly alarm method according to claim 6, characterized in that, The specific steps of S5 are as follows: S501: Extract the abnormal target number based on the intrusion anomaly verification result, call the trajectory node data corresponding to the target number, filter the zone division identifier parameters attached to the node, summarize the correspondence between the target number and the zone identifier, and generate alarm zone mapping value. S502: Based on the alarm zone mapping value, extract the trajectory dwell time of the target number in the zone, the zone entry frequency and the trigger node density, match and judge against the alarm trigger condition value configured in the zone, filter the target numbers that meet the conditions, and generate alarm trigger matching results; S503: Extract the target number and the identification of the zone according to the alarm trigger matching result, construct the alarm command identification and output it to the intrusion anomaly alarm interface, and record the interface response time and alarm level identification to generate the area intrusion anomaly alarm result.

8. The regional intrusion anomaly alarm method according to claim 1, characterized in that, The multiple sensing agents refer to sensing nodes deployed within the protected area that are capable of independently collecting the spatial coordinates of moving targets and generating corresponding timestamps; The protected area refers to the monitoring space range pre-set by the regional intrusion anomaly alarm system, and the range is defined by the regional boundary division or the protection zone identification; The spatial position coordinates refer to the coordinate data used to represent the position of a moving target within the protected area, which is generated based on a unified coordinate system; The timestamp refers to the time stamp generated when the sensing agent collects the corresponding spatial location coordinates; The normal direction change threshold refers to the direction change reference limit preset by the system, which is used to limit the range of direction changes of a moving target in the normal travel state within the protected area.

9. The regional intrusion anomaly alarm method according to claim 1, characterized in that, The zone division identifier refers to the zone identifier information used to distinguish different monitoring sub-regions within the protected area and is associated with the trajectory node; The so-called stationary state refers to a state in which a moving target is continuously in a positionally restricted state within the same defense zone; The alarm zone identifier refers to the area identifier in the zone division identifier that is associated with the alarm triggering rule.

10. A multi-agent fusion perception regional intrusion anomaly alarm system, characterized in that, The system is used to implement the regional intrusion anomaly alarm method according to any one of claims 1-9.

Images