Method and system for identifying and warning abnormal events based on campus multi-source monitoring data

Abstract

The application discloses a campus multi-source monitoring data-based abnormal event identification and early warning method and system, and relates to the technical field of safety management.The campus multi-source monitoring data-based abnormal event identification and early warning method comprises the following steps: S1, collecting campus multi-source monitoring data, and performing preprocessing operation on the campus multi-source monitoring data; S2, reading a gate area passing detection chain and a historical passing sample atlas library, and performing passing reverse analysis; S3, calling reverse checking records, card swiping identity sequence chains and target crossing sequence chains to perform crossing attribution analysis; and S4, combining crossing attribution result tables, trailing candidate records, supplementary swiping passing records, trailing stripping records and attribution records to be checked, performing trailing alarm closed verification, and generating a trailing early warning report, a passing trace record and a trailing closed insufficient retention record.The method solves the problems of trailing abnormal misjudgment and incomplete evidence caused by the dislocation of campus gate area passing records.

Description

Technical Field

[0001] This invention belongs to the field of safety management technology, specifically a method and system for identifying and warning of abnormal events based on multi-source monitoring data on campus. Background Technology

[0002] In campus security management, access control devices, magnetic door sensors, door lock controllers, and door area cameras are widely used for entrance and exit access management. Access control devices typically record card swipe identity, swipe time, and swipe result; magnetic door sensors and door lock controllers record door opening and closing events; and door area cameras capture images or video clips of people passing through the door line. Existing technologies include early warning systems for campus security, such as those using radar detection, video surveillance, smoke detection, and alarm management, or those using campus monitoring and security equipment to collect information and analyze student behavior to issue alarm notifications. These solutions can improve the ability to detect campus security incidents, but their focus is mostly on gatherings of people, abnormal smoke, abnormal student behavior, or comprehensive security management. They are still insufficient to address issues such as door opening retention, delayed swipes, waiting while holding the door, and tailgating errors caused by multiple targets continuously crossing the door line in access control scenarios.

[0003] For example, invention patent CN118135770B discloses a campus security prevention and early warning system and method. The system includes: a radar detection module for indoor human detection on campus to obtain information on personnel gathering; and an abnormal gathering alarm signal based on abnormal personnel gathering; a video monitoring module for recording facial images of gathered personnel based on the abnormal gathering alarm signal; a smoke detection module for detecting smoke particles in the campus air using a smoke detector and issuing an abnormal smoke alarm signal; a display module for creating a 3D model of campus rooms; and a functional module for equipment management, alarm management, and floor management, thus completing the campus security prevention and early warning system. This provides comprehensive protection for campus safety and bullying prevention.

[0004] For example, the invention patent with announcement number CN116895128B discloses a comprehensive campus behavior early warning system, including: an information receiving module, which is used to collect various information on campus through monitoring and security equipment; a behavior analysis module, which is used to analyze and judge student behavior in campus videos collected by the information receiving module and issue alarm signals; a hidden danger early warning module, which comprehensively analyzes student behavior and actions and issues alarm signals; and an alarm notification module, which sends a notification to management personnel after receiving an alarm signal. This comprehensive system can establish a new front-end deployment structure for integrated data analysis and abnormal behavior security management, while also being compatible with existing equipment and enabling intelligent transformation of existing security equipment to provide a more efficient and comprehensive management system.

[0005] However, in campus access control scenarios, doors often remain open for a period of time after a card is swiped, allowing multiple students or individuals to pass through the same door opening area sequentially. During passage, situations may arise such as pushing the door open before swiping again, holding onto the door while waiting, obstructing passage, multiple people passing close together, and delayed writing of card swipe records. Existing tailgating detection methods, if relying solely on a single card swipe record and binding it to a single target passing through the door, are prone to misjudging normal follow-up swipes as tailgating, or potentially classifying the actual tailgating target as a previous legitimate card swipe record, leading to discrepancies between the alarm target and the actual person passing through.

[0006] Furthermore, existing access control anomaly identification methods, when processing multi-source data, typically lack unified verification of door opening intervals, card swipe sequence, door cable crossing order, and video target appearance order. Access control logs, door magnetic sensor opening / closing records, and video passage records differ in collection frequency, recording granularity, and timestamp sources. Without effective temporal correlation and evidence closure processing, this can easily lead to inaccurate anomaly origin location, misattribution of trailing objects, incomplete alarm record evidence, and increased difficulty in subsequent tracing and evidence collection.

[0007] Therefore, in order to address the above issues, there is an urgent need for an abnormal event identification and early warning method and system based on multi-source monitoring data on campus. Summary of the Invention

[0008] To address the above problems, this invention provides an abnormal event identification and early warning method and system based on multi-source campus monitoring data, which solves the problems of misjudgment of tailgating anomalies and incomplete evidence collection caused by misaligned access records at campus gates.

[0009] To achieve the above objectives, the technical solution adopted by this invention is: an abnormal event identification and early warning method based on campus multi-source monitoring data, including S1, collecting campus multi-source monitoring data and performing preprocessing operations on the campus multi-source monitoring data; S2, reading the gate area access detection chain and historical access sample map library, performing access reversal analysis, and generating access sequence passage records and reversal verification records based on the access reversal analysis results; S3, calling the reversal verification records, card swipe identity sequence chain and target crossing sequence chain to perform crossing attribution analysis, and performing supplementary swipe attribution, tailgating candidate identification and object misbinding removal based on the crossing attribution analysis results; S4, combining the crossing attribution result table, tailgating candidate records, supplementary access records, tailgating removal records and attribution pending verification records to perform tailgating alarm closure verification, and generating tailgating early warning reports, access tracing records and tailgating closure insufficiency retention records based on the tailgating alarm closure verification results.

[0010] Furthermore, the specific steps for collecting multi-source monitoring data on campus are as follows: Collecting multi-source monitoring data on campus: Obtaining the access control point number, card swipe record number, card swipe identity identifier, card swipe time, card swipe result code, and access record number through the card swipe log interface of the access control controller; Obtaining the door number, opening event time, closing event time, door open duration, and door event code through the opening and closing event logs of the door magnetic sensor and door lock controller; Obtaining the video channel number, camera number, and door cable number through the video stream of the door area camera and the door cable detection program. The data includes: video target number, target appearance time, door push action time, door line crossing time, crossing direction marker, door holding waiting marker, and video evidence segment number; access control point number, door number, camera number, door line number, door line position, and door passage direction obtained from the access control point configuration table; and normal swipe sample markers, tailgating sample markers, door holding waiting sample markers, sample card swipe trigger point, sample door opening point, sample door push action point, sample target appearance point, sample door line crossing point, and sample door closing point obtained from historical passage samples.

[0011] Furthermore, the specific steps for preprocessing campus multi-source monitoring data are as follows: Perform time calibration, duplicate swipe merging, and failed swipe isolation archiving on access control point numbers, card swipe record numbers, card swipe identification, card swipe time, and card swipe result codes; generate a card swipe identity sequence chain according to the card swipe time; perform opening interval slicing on door numbers, door opening event times, door closing event times, door opening duration, and door event codes to generate door opening interval numbers and door opening intervals; process video channel numbers, camera numbers, and door cable codes... The system performs door line mapping, trajectory capture, and crossing positioning based on the following data: access control point number, door number, camera number, and door line number; door area access mapping table; and writes the card swipe sequence chain, target crossing sequence chain, door area access mapping table, and historical access sample data into the door area access detection chain, using the door opening interval number as the linking index.

[0012] Furthermore, the specific steps for performing passage reversal analysis by reading the gate area access detection chain and historical access sample graph library are as follows: Read the gate area access detection chain, using the gate opening interval number as an index, and combine the card swipe time, door opening event time, door pushing action time, target appearance time, door line crossing time, and door closing event time to construct a gate event sequence graph. The gate event sequence graph includes a set of graph points and a set of directed edges. The set of graph points includes the card swipe trigger point, door opening point, door pushing action point, target appearance point, door line crossing point, and door closing point. The set of directed edges includes directed edges indicating the sequence of events and the gate opening interval affiliation. The system reads the historical passage sample graph library, which consists of sample card swipe trigger points, sample door opening points, sample door pushing action points, sample target appearance points, sample door line crossing points, sample door closing points, and directed edges representing the sequence of sample events. Using graph sequence dynamic regularization technology, the system aligns the time sequence of the door event sequence graph with the sample graph in the historical passage sample graph library and matches the directed edge paths. Based on the differences in the time sequence of the graph points and the differences in the directed edge paths, it identifies follow-up swipe paths, non-swipe card continuation crossing paths, door holding waiting paths, and closing conflict paths, and generates passage reversal values.

[0013] Furthermore, the specific steps for generating passage sequence records and reversal verification records based on the passage reversal analysis results are as follows: compare the passage reversal value and the reversal threshold. When the passage reversal value is less than the reversal threshold, bind the card swipe record and the door crossing record according to the card swipe first rule to generate the passage sequence record; when the passage reversal value is greater than or equal to the reversal threshold, the one-to-one binding relationship between a single card swipe record and a single door crossing target is not accepted for the time being, and the card swipe identity sequence chain, target crossing sequence chain, door event sequence diagram, supplementary swipe post-path, no-card-swipe continuation crossing path, door holding waiting path, and closed conflict path are written into the reversal verification record.

[0014] Furthermore, the specific steps for performing crossing attribution analysis by calling the reverse verification record, card swipe identity sequence chain, and target crossing sequence chain are as follows: Read the passage sequence record, reverse verification record, card swipe identity sequence chain, target crossing sequence chain, and door event sequence diagram. Using the door opening interval number as an index, arrange the card swipe record numbers according to the card swipe time to generate an interval card swipe identity queue. Arrange the video target numbers according to the door line crossing time to generate an interval target crossing queue. Read the door line crossing sequence number and video evidence segment number from the target crossing sequence chain. Read the supplementary swipe path, no-card-swipe continuation crossing path, door-holding waiting path, and closed conflict path from the reverse verification record. Then, combine the interval card swipe identity queue, interval target crossing queue, supplementary swipe path, no-card-swipe continuation crossing path, door-holding waiting path, and closed conflict path. The conflicting paths are written into the crossing attribution combination table; a card-swiping crossing bipartite graph is constructed using the interval card-swiping identity queue as the card-swiping side node and the interval target crossing queue as the target side node; the assignment cost between the card-swiping side node and the target side node is generated based on the time difference between the card-swiping time and the door line crossing time, the crossing direction marker, the re-swiping post-path, the door-holding waiting path, and the closed conflicting path; with the goal of minimizing the total assignment cost, a card-swiping crossing attribution assignment model is constructed by combining the crossing attribution combination table and the door event sequence graph, generating a crossing attribution confidence value and a crossing attribution result table; the crossing attribution result table includes authorized attribution marker, re-swiping attribution marker, tailing candidate marker, object misbinding marker, door-holding waiting marker, card-swiping record number, card-swiping identity identifier, video target number, door line crossing sequence number, and video evidence segment number.

[0015] Furthermore, the specific steps for performing supplementary attribution, tailgating candidate identification, and object misbinding removal based on the crossing attribution analysis results are as follows: Compare the crossing attribution confidence value with the confidence threshold. When the crossing attribution confidence value is greater than or equal to the confidence threshold, the crossing attribution result table is accepted. Records with supplementary attribution markers are written into the supplementary passage record, and the door opening interval number and video target number corresponding to records with tailgating candidate markers are written into the tailgating candidate record. The video target number with object misbinding markers is disconnected from the previous authorized passage record and written into the tailgating removal record. The tailgating removal record carries the previous authorized passage record number, the removed candidate object record number, and the video evidence segment number. When the crossing attribution confidence value is less than the confidence threshold, the updating of the supplementary passage record and the tailgating removal record is paused. The corresponding video target number is written into the attribution pending verification record, and the crossing attribution combination table and the video evidence segment number are retained for evidence verification.

[0016] Furthermore, combining the crossing attribution result table, tailgating candidate records, re-swiping records, tailgating stripping records, and attribution pending verification records, the specific steps for tailgating alarm closure verification are as follows: Read the crossing attribution result table, re-swiping records, tailgating candidate records, tailgating stripping records, attribution pending verification records, and video evidence segment numbers; form a candidate object row index using the door opening interval number in the tailgating candidate records and tailgating stripping records and the video target number; generate a door line crossing evidence bit based on the door line crossing sequence number and video evidence segment number in the tailgating candidate records; generate a card swipe gap evidence bit based on records in the crossing attribution result table that neither hit the authorized attribution mark nor the re-swiping attribution mark; generate a re-swiping exclusion evidence bit when there is no record in the re-swiping records that matches the video target number of the tailgating candidate records; and generate a re-swiping exclusion evidence bit based on the previous... An object stripping evidence bit is generated from an authorized access record number, a candidate object record number after stripping, and a video evidence segment number. When there is a record in the pending record that matches the video target number of the candidate object row, an evidence collection suspension and suppression bit is generated. An evidence column index is formed using the door line crossing evidence bit, card swipe gap evidence bit, re-swipe exclusion evidence bit, object stripping evidence bit, and evidence collection suspension and suppression bit, and a trailing evidence closure matrix is ​​constructed. According to the candidate object row index and the evidence column index, the matrix elements that hit each evidence bit are assigned a value of one, and the matrix elements that do not hit are assigned a value of zero. A trailing closure weight vector is generated according to the evidence column index, where the evidence collection suspension and suppression bit corresponds to a negative weight. The trailing evidence closure matrix and the trailing closure weight vector are multiplied to obtain the candidate object closure vector, and the vector components of the candidate object row are used as the trailing alarm closure value.

[0017] Furthermore, the specific steps for generating a tailgating warning report, passage tracing record, and tailgating closure insufficiency retention record based on the tailgating alarm closure verification result are as follows: compare the tailgating alarm closure value with the closure threshold. When the tailgating alarm closure value is greater than or equal to the closure threshold, generate a tailgating warning report, mark the video target number as a suspected tailgating object, and write the previous authorized passage record number, the candidate object record number after stripping, and the video evidence segment number into the passage tracing record. Write the door line crossing record, card swipe gap record, supplementary swipe exclusion record, and object stripping record associated with the evidence bit into the monitoring log. When the tailgating alarm closure value is less than the closure threshold, write the corresponding video target number into the tailgating closure insufficiency retention record, retain the tailgating evidence closure matrix, the crossing attribution combination table, and the video evidence segment number, and do not generate a tailgating warning report for the time being.

[0018] Furthermore, the second aspect of the present invention provides an abnormal event identification and early warning system based on campus multi-source monitoring data, and applies an abnormal event identification and early warning method based on campus multi-source monitoring data, including: a data acquisition and preprocessing module, used to acquire campus multi-source monitoring data and perform preprocessing operations on the campus multi-source monitoring data; a door opening and holding reversal discrimination module, used to read the door area access detection chain and historical access sample map library, perform access reversal analysis, and generate access sequence passage records and reversal verification records based on the access reversal analysis results; a swipe attribution and tailgating object stripping module, used to call the reversal verification record, card swipe identity sequence chain and target crossing sequence chain to perform crossing attribution analysis, and perform swipe attribution, tailgating candidate identification and object misbinding stripping based on the crossing attribution analysis results; and a tailgating evidence closure and early warning output module, used to combine the crossing attribution result table, tailgating candidate records, swipe access records, tailgating stripping records and attribution pending verification records to perform tailgating alarm closure verification, and generate tailgating early warning reports, access tracing records and tailgating closure insufficiency retention records based on the tailgating alarm closure verification results.

[0019] The present invention has the following beneficial effects:

[0020] (1) This invention collects multi-source monitoring data from the campus and generates a card swipe identity sequence chain, a target crossing sequence chain, a door area access mapping table, and a door area access detection chain. This enables access control point numbers, card swipe record numbers, card swipe times, door opening interval numbers, door line crossing times, and video target numbers to establish associations around the same door opening interval number. This achieves the effect of collaborative verification of card swipe identity sequence, door opening interval, and target crossing sequence, effectively solving the problem of incomplete access attribution basis caused by the scattered processing of access control logs, door magnetic opening and closing records, and video door passage records in the prior art.

[0021] (2) This invention constructs a door event sequence diagram by reading the door area access detection chain and historical access sample map library, and identifies the post-swiping path, the non-swiping continuous crossing path, the door holding waiting path and the closing conflict path. This enables the card swiping trigger point, door opening point, door pushing action point, target appearance point, door line crossing point and door closing point within the same door opening interval to be checked according to the order of events. This achieves the effect of access reversal recognition in the door opening and holding scenario, and effectively solves the problem that the prior art is easily misjudged as tailing when pushing the door first and then swiping, holding the door waiting and multiple targets continuously crossing the door line.

[0022] (3) This invention performs crossing attribution analysis by calling the reverse verification record, the card swipe identity sequence chain and the target crossing sequence chain, and generates a crossing attribution result table, a supplementary swipe passage record, a tailing candidate record and a tailing stripping record, so that a traceable attribution relationship is formed between the card swipe record number, the card swipe identity identifier, the video target number, the door line crossing sequence number and the video evidence segment number, thereby achieving the effects of supplementary swipe attribution, tailing candidate identification and object misbinding stripping, effectively solving the problem in the prior art that the real tailing target is assigned to the previous authorized passage record, causing the disposal object to be offset.

[0023] (4) This invention combines the crossing attribution result table, supplementary passage record, tailing candidate record, tailing stripping record and attribution pending verification record to perform tailing alarm closure verification, and generates tailing warning report, passage tracing record and tailing closure insufficiency retention record, so that the door line crossing evidence bit, card swipe gap evidence bit, supplementary swipe exclusion evidence bit, object stripping evidence bit and evidence collection postponement suppression bit can participate in the tailing alarm closure judgment, thereby achieving the effect of tailing warning evidence closure and tracing evidence collection retention, effectively solving the problems of incomplete tailing alarm evidence chain and subsequent evidence collection and verification difficulties in the prior art. Attached Figure Description

[0024] Figure 1 This is a flowchart of the abnormal event identification and early warning method based on multi-source campus monitoring data according to the present invention;

[0025] Figure 2 This is a structural diagram of the abnormal event identification and early warning system based on multi-source campus monitoring data of the present invention;

[0026] Figure 3 This is a data diagram showing the passage reversal of the door opening section based on the supplementary swipe path and the non-swipe card connection passage path of the present invention;

[0027] Figure 4 This is a schematic diagram of the trailing evidence closed matrix based on the candidate object row index and the evidence column index of the present invention. Detailed Implementation

[0028] To enable those skilled in the art to better understand the technical solution, the present invention will be described in detail below with reference to embodiments. The description in this part is only exemplary and explanatory, and should not be used to limit the scope of protection of the present invention in any way.

[0029] Please see Figures 1-4This invention provides a technical solution: an abnormal event identification and early warning method based on campus multi-source monitoring data, including S1, collecting campus multi-source monitoring data and performing preprocessing operations on the campus multi-source monitoring data; S2, reading the gate area access detection chain and historical access sample map library, performing access reversal analysis, and generating access sequence records and reversal verification records based on the access reversal analysis results; S3, calling the reversal verification records, card swipe identity sequence chain and target crossing sequence chain to perform crossing attribution analysis, and performing supplementary swipe attribution, tailgating candidate identification and object misbinding removal based on the crossing attribution analysis results; S4, combining the crossing attribution result table, tailgating candidate records, supplementary access records, tailgating removal records and attribution pending verification records to perform tailgating alarm closure verification, and generating tailgating early warning reports, access tracing records and tailgating closure insufficiency retention records based on the tailgating alarm closure verification results.

[0030] Specifically, the steps for collecting multi-source monitoring data on campus are as follows: Collecting multi-source monitoring data on campus: Obtaining the access control point number, card swipe record number, card swipe identity identifier, card swipe time, card swipe result code, and access record number through the card swipe log interface of the access control controller; obtaining the door number, door opening event time, door closing event time, door opening duration, and door event code through the opening and closing event logs of the door magnetic sensor and door lock controller; obtaining the video channel number, camera number, door line number, video target number, target appearance time, and push notification information through the video stream of the door area camera and the door line detection program. The data includes the door action time, door line crossing time, crossing direction marker, door-holding waiting marker, and video evidence segment number. The door-holding waiting marker is generated based on the video target number's dwell state in the door line location neighborhood, the door-pushing action time, the door event code, and the subsequent target crossing relationship. When a video target number is located in the door line location neighborhood within the same door opening interval, the door-pushing action time is earlier than the door line crossing time of the subsequent video target number, the door event code continuously indicates the door is open, and the video target number maintains a door-holding or waiting posture for at least one second before the subsequent video target number completes its door line crossing, the data is considered valid. Write the corresponding video target number into the door-holding waiting marker; obtain the access control point number, door number, camera number, door line number, door line position, and door passage direction through the access control point configuration table; obtain normal re-swipe sample markers, tailgating sample markers, door-holding waiting sample markers, sample card swipe trigger points, sample door opening points, sample door pushing action points, sample target appearance points, sample door line crossing points, and sample door closing points through historical passage samples. The historical passage samples consist of archived card swipe logs, opening and closing event logs, video evidence clips, passage traceability records, and management terminal review records. Together, they form the following: a normal supplementary swipe sample marker is generated based on the swipe result code indicating passage, the swipe time being later than the corresponding door line crossing time, and the passage tracing record confirming that the swipe identity identifier corresponds to the video target number; a trailing passage sample marker is generated based on the existence of a video target number within the same door opening interval that does not match the swipe identity identifier and is confirmed by the passage tracing record as a continuous crossing; and a door-holding waiting sample marker is generated based on the historical record of the video target number maintaining a door-holding or waiting posture within the door line position neighborhood, the door event code continuously indicating an open state, and the subsequent video target number completing door line crossing.

[0031] This implementation plan unifies the collection of card swipe logs, door opening and closing events, door area video streams, door line detection results, access control point configuration tables, and historical passage samples. This enables access control point numbers, door numbers, camera numbers, and door line numbers to form a data correspondence within the same door area. Furthermore, it reconstructs the personnel passage process by recording the card swipe time, door opening event time, target appearance time, door pushing action time, door line crossing time, and door closing event time. Simultaneously, it supplements the status information under different passage scenarios by using door-holding waiting markers, normal swipe sample markers, tailgating sample markers, and door-holding waiting sample markers. This provides a complete and traceable data foundation for subsequent passage reversal analysis, crossing attribution analysis, swipe attribution judgment, and tailgating candidate identification, thereby reducing the risk of misidentification due to different sources among card swipe records, door opening records, and video target crossing records.

[0032] Specifically, the preprocessing steps for multi-source monitoring data on campus are as follows: Time calibration, duplicate swipe merging, and failed swipe isolation and archiving are performed on the access control point number, card swipe record number, card swipe identification, card swipe time, and card swipe result code. Time calibration uses the access control controller server clock as a unified time reference, converting the card swipe time to a time series under the unified time reference. Duplicate swipe merging uses the access control point number, card swipe identification, and card swipe result code as the merging key. When the swipe time interval of multiple card swipe record numbers under the same merging key is no greater than three seconds, the card swipe record number corresponding to the earliest swipe time is retained, and the remaining swipe record numbers are written into the duplicate swipe merging record. Failed swipe isolation and archiving uses the card swipe result code... Using unauthorized access, invalid card number, timeout, or denied access as identification criteria, the corresponding card swipe record number is written into the failed card swipe isolation archive record, and a card swipe identity sequence chain is generated according to the card swipe time. Opening interval slicing is performed on the door number, opening event time, closing event time, door opening duration, and door event code. The opening event corresponding to the door event code is used as the starting point of the door opening interval, and the first closing event following the opening event under the same door number is used as the ending point of the door opening interval, generating a door opening interval number and a door opening interval. When multiple consecutive opening events occur under the same door number without a closing event, the earliest opening event time is retained as the starting point, and the remaining opening events are written into the abnormal opening / closing event record. Recording; when a closing event is missing after an opening event, the closing time is the time after the door is open and held open, plus the time after the door is open, and the corresponding door event code is written to the abnormal opening / closing event record; when a closing event lacks a corresponding opening event, the closing event is written to the abnormal opening / closing event record, and no door opening interval is generated; adjacent door opening intervals are divided by the closing event time and the next opening event time, and different door opening interval numbers are generated respectively; door line mapping, trajectory interception, and crossing positioning are performed on the video channel number, camera number, door line number, video target number, target appearance time, door pushing action time, door line crossing time, crossing direction mark, door holding waiting mark, and video evidence segment number. Door line mapping is based on... The target trajectory segment is associated with the corresponding door line number based on the camera number, door line number, and door line position; trajectory extraction takes the video target number within the same door opening range as the object, extracts the target trajectory segment between the time the target appears and the time the door closes, and retains the corresponding video evidence segment number; crossing positioning takes the first video frame when the target trajectory segment intersects with the door line position as the door line crossing time, generates a crossing direction mark based on the change direction of the target trajectory segment relative to the door line position before and after the intersection, takes the first video frame when the door pushing action appears in the target trajectory segment as the door pushing action time, generates a door holding waiting mark based on the record of the target trajectory segment staying near the door line position without completing the crossing, and generates a target crossing sequence chain;Map the access control point numbers, door numbers, camera numbers, and door cable numbers to generate a door area access mapping table. Using the door opening interval number as the linking index, write the card swipe sequence chain, target passage sequence chain, door area access mapping table, and historical access sample data into the door area access detection chain.

[0033] In this implementation plan, by uniformly calibrating the access control point numbers, card swipe record numbers, card swipe identification, card swipe time, and card swipe result codes, merging duplicate swipes, and isolating and archiving failed swipes, card swipe records can form a sequential chain of card swipe identities based on a unified time benchmark. By slicing the door number, door opening event time, door closing event time, door opening duration, and door event code into opening interval segments, the door opening and closing process is divided into door opening interval numbers and door opening intervals with clearly defined start and end boundaries. Furthermore, by analyzing the video channel number, camera number, door cable number, video target number, target appearance time, door pushing action time, door cable crossing time, crossing direction marking, and door holding time... The system uses markers and video evidence segment numbers for door line mapping, trajectory capture, and crossing location to generate a target crossing sequence chain during the video target's passage through the door. Then, it maps the access control point number, door number, camera number, and door line number to the card swipe identity sequence chain, target crossing sequence chain, door area access mapping table, and historical access sample data into the door area access detection chain. This achieves unified linking and temporal organization between card swipe records, door opening / closing records, video passage records, and historical access samples, effectively solving the problems of inconsistent time bases for multi-source access control data, unclear door opening interval boundaries, and difficulty in accurately associating video target crossing events with card swipe records in existing technologies.

[0034] Specifically, the steps for performing passage reversal analysis by reading the gate area access detection chain and historical access sample graph database are as follows: Read the gate area access detection chain, using the gate opening interval number as an index, and combine the card swipe time, door opening event time, door pushing action time, target appearance time, door line crossing time, and door closing event time to construct a gate event sequence graph. The gate event sequence graph includes a set of graph points and a set of directed edges. The set of graph points includes the card swipe trigger point, door opening point, door pushing action point, target appearance point, door line crossing point, and door closing point. The set of directed edges includes directed edges indicating the order of events and the direction of the gate opening interval. The directed edges indicating the order of events represent the sequential relationship between graph points within the same gate opening interval according to the event times. The directed edges indicating the attribution of each graph point to the gate opening interval number are used to represent the attribution relationship between each graph point and the gate opening interval number. The historical access sample graph database is read. The sample graphs in the historical access sample graph database consist of sample card swipe trigger points, sample gate opening points, sample door push action points, sample target appearance points, sample door line crossing points, sample gate closing points, and directed edges indicating the sequence of sample events. Directed edges indicating the attribution of sample gate opening intervals are added according to the sample gate opening intervals to which the sample graph points belong, ensuring that the gate event sequence graph and the sample graphs maintain a matching relationship in terms of graph point type, event sequence directed edges, and gate opening interval attribution directed edges. Through graph sequence dynamic normalization technology, the graph points in the gate event sequence graph and the sample graphs in the historical access sample graph database are compared. The process involves sequence alignment and directed edge path matching. The graph sequence dynamic regularization technique uses a set of graph points ordered by event time as the sequence of graph points to be matched, and a set of sample graph points ordered by sample event time as the sample sequence of graph points. The time difference between graph points of the same type, normalized by the duration of door opening, is used as the graph point temporal distance. The difference in directed edge path distance is calculated based on the missing, reversed, or cross-door opening intervals of directed edges belonging to the corresponding directed edges. The temporal distance and directed edge path distance are accumulated according to the regularized path with monotonically increasing event time, and the regularized path with the smallest accumulated distance is selected as the alignment result. Based on the differences in graph point temporal distance and directed edge path, the technique identifies subsequent swipe paths and non-swipe card continuation traversal paths. The document defines the following paths for door waiting and closing conflict paths: a supplementary swipe path (where the door line crossing point is earlier than the corresponding card swipe trigger point and the corresponding card swipe result code indicates passage), a non-card swipe continuation crossing path (where the subsequent door line crossing point within the same door opening range lacks a matching card swipe trigger point and continues from the previous door line crossing point), a door-holding waiting path (where the door pushing action point is earlier than the subsequent door line crossing point and corresponds to the door-holding waiting marker), and a closing conflict path (where the door closing point is earlier than the door line crossing point or the door line crossing point falls outside the directed edge of the door opening range). A passage reversal value is generated based on the differences in point timing, directed edge path differences, and the hit results of supplementary swipe paths, non-card swipe continuation crossing paths, door-holding waiting paths, and closing conflict paths. Figure 3The diagram shows the passage reversal data of the door opening section based on the supplementary swipe post-path and the non-swipe continuation traversal path in this embodiment. The diagram uses swipe record 1, video target 1, video target 2, and swipe record 2 as passage record rows. The swipe time, door opening event time, target appearance time, door pushing action time, door line crossing time, and door closing event time are displayed according to the sequence of events within the same door opening section. Among them, video target 1 completes the door line crossing first, and video target 2 completes the door line crossing in the same door opening section. The swipe time of swipe record 2 is after the door line crossing time of video target 2, forming the supplementary swipe post-path. Combined with the continuation traversal relationship between video target 1 and video target 2, a non-swipe continuation traversal path is formed. This intuitively shows the passage reversal situation between the swipe identity sequence chain and the target crossing sequence chain, providing data basis for subsequent reversal verification record generation, supplementary swipe attribution analysis, and tailgating candidate identification. The passage reversal value is a weighted sum of graph point temporal differences, directed edge path differences, and path hit results, truncated to a range of zero to one. The weights of each sub-cost are obtained based on the contribution ratio of time difference, direction consistency, post-refresh, gate waiting, and closure conflict to the crossing attribution verification results in historical passage samples, and the sum of all weights is one. The larger the passage reversal value, the greater the deviation of the gate event sequence graph from the normal passage order in the historical passage sample graph library.

[0035] In this implementation plan, the card swipe time, door opening event time, door pushing action time, target appearance time, door line crossing time, and door closing event time are organized into a door event sequence diagram by using the door opening interval number as an index. The sample diagrams in the historical passage sample diagram library are supplemented to have the same point type and directed edge type, so that the door event sequence diagram and the sample diagram can be aligned in the same structure for point timing and directed edge path matching. By generating passage reversal values ​​through point timing differences, directed edge path differences, and closed conflict path penalty terms, passage reversal values ​​can be generated, which can distinguish between the supplementary swipe path, the no-card-swipe continuation passage path, the door holding waiting path, and the closed conflict path. This provides a quantifiable basis for the subsequent generation of passage sequence records and reversal verification records, reducing the risk that normal supplementary swipes during the door opening period are misjudged as tailgating, and that real tailgating is classified into the previous authorized passage record.

[0036] Specifically, the steps for generating passage sequence records and reversal verification records based on the passage reversal analysis results are as follows: Compare the passage reversal value and the reversal threshold. The reversal threshold is obtained from the distribution of passage reversal values ​​corresponding to normal swipe sample markers, trailing passage sample markers, and door-holding waiting sample markers in the historical passage sample map library. When the passage reversal value is less than the reversal threshold, bind the card swipe record and the door line crossing record according to the card-swipe-first rule to generate a passage sequence record. The card-swipe-first rule is as follows: within the same door opening range, the card swipe result code indicates passage, the card swipe time is earlier than the corresponding door line crossing time, and the time difference between the card swipe time and the corresponding door line crossing time is not greater than the allowed card-swipe crossing time difference. The crossing direction marker is consistent with the door passage direction. The allowed card-swipe crossing time difference is based on the sample swipe values ​​in the historical passage samples of the same access control point. The time difference distribution from the card trigger point to the sample door line crossing point is obtained and is less than or equal to the door opening duration. When there are multiple card swipe records and multiple video target numbers within the same door opening interval, the door line crossing records in the target crossing sequence chain are matched sequentially according to the card swipe identity sequence chain. Only one card swipe record is allowed to be bound to the same door line crossing record. When there are multiple bindable objects, the record with the smallest time difference between the card swipe time and the door line crossing time and which has not yet been bound is selected to generate the passage sequence record. When the passage reversal value is greater than or equal to the reversal threshold, the one-to-one binding relationship between a single card swipe record and a single door crossing target is not accepted. The card swipe identity sequence chain, target crossing sequence chain, door event sequence diagram, supplementary swipe post-path, no-card-swipe continuation crossing path, door-holding waiting path, and closing conflict path are written into the reversal verification record.

[0037] In this implementation plan, normal passage and suspected reverse passage within the gate opening section are separated by comparing the passage reversal value and the reversal threshold. When the passage reversal value is less than the reversal threshold, the card swiping time is confirmed to be earlier than the gate crossing time, the time difference meets the allowed time difference for card crossing, and the crossing direction mark is consistent with the gate passage direction, based on the card swiping priority rule. The card swiping record is then bound to the gate crossing record according to the card swiping identity sequence chain and the target crossing sequence chain to generate a passage sequence record. When the passage reversal value is greater than or equal to the reversal threshold, the one-to-one binding relationship between a single card swiping record and a single gate crossing target is not adopted. Instead, the card swiping identity sequence chain, the target crossing sequence chain, the gate event sequence diagram, the supplementary swiping post-path, the no-card-swiping continuation crossing path, the door-holding waiting path, and the closed conflict path are written into the reversal verification record. This avoids incorrect binding in supplementary swiping, continuation crossing, and door-holding waiting scenarios, and provides a more reliable verification basis for subsequent crossing attribution analysis, supplementary swiping attribution judgment, and tailgating candidate identification.

[0038] Specifically, the steps for performing crossing attribution analysis by calling the reverse verification record, card swipe identity sequence chain, and target crossing sequence chain are as follows: Read the passage sequence record, reverse verification record, card swipe identity sequence chain, target crossing sequence chain, and door event sequence diagram. Using the door opening interval number as an index, arrange the card swipe record numbers according to the card swipe time to generate an interval card swipe identity queue. Arrange the video target numbers according to the door line crossing time to generate an interval target crossing queue. Read the door line crossing sequence number and video evidence segment number from the target crossing sequence chain. Read the supplementary swipe path, no-card-swipe continuation crossing path, door-holding waiting path, and closed conflict path from the reverse verification record. Then, combine the interval card swipe identity queue, the interval... The target crossing queue, the follow-up path after card swiping, the no-card-swiping continuation crossing path, the door-holding waiting path, and the closed conflict path are written into the crossing attribution combination table. Using the interval card-swiping identity queue as the card-swiping side node and the interval target crossing queue as the target side node, a card-swiping crossing bipartite graph is constructed. Each card-swiping side node connects to at most one target side node, and each target side node receives at most one card-swiping side node. The assignment cost between the card-swiping side node and the target side node is generated based on the time difference between the card swiping time and the door-line crossing time, the crossing direction marker, the follow-up path after card swiping, the door-holding waiting path, and the closed conflict path. The time difference sub-cost is calculated by dividing the absolute amount of the difference between the card swiping time and the door-line crossing time by the door opening time. The duration is determined by the following conditions: If the door opening duration is exceeded, the time difference sub-cost is set to one; the direction consistency sub-cost is determined based on whether the crossing direction marker matches the door opening interval affiliation in the door event sequence diagram; it is set to zero if it matches, and one if it doesn't; the post-swiping sub-cost is determined based on the post-swiping path; it is set to zero if the door line crossing time of the target node is earlier than the card swiping time of the card-swiping node and they are within the same door opening interval number, otherwise it is set to one; the door-holding waiting sub-cost is determined based on the door-holding waiting path; it is set to one if the target node corresponds to the door-holding waiting path and no independent door line crossing record is formed; the closure conflict sub-cost is determined based on the closure conflict path; when the target node's door line crosses... When the time is later than the closing event or does not belong to the corresponding door opening interval number, the assignment edge between the card-swiping side node and the target side node is removed. The assignment cost is obtained by multiplying the time difference sub-cost, the direction consistency sub-cost, the supplementary swipe post-sub-cost, the door holding waiting sub-cost, and the closure conflict sub-cost by their respective weights and then summing them up. Each sub-cost is limited to between zero and one. With the goal of minimizing the total assignment cost, a card-swiping crossing assignment model is constructed by combining the crossing attribution combination table and the door event sequence diagram. When the number of card-swiping side nodes and the number of target side nodes are inconsistent, the target side node without a valid assignment edge is retained as an unmatched target side node, and the card-swiping side node without a target side node to match is retained as an unmatched card-swiping side node.When multiple candidate assignments have the same total cost, the candidate assignment with the smallest sum of time difference sub-costs is selected first. If they are still the same, the candidate assignment with the earlier door line crossing sequence number and zero closing conflict sub-cost is selected first. A crossing attribution confidence value is generated based on the difference between the total cost of the optimal assignment and the second-best assignment. The crossing attribution confidence value ranges from zero to one. The smaller the total cost and the larger the difference between the second-best and optimal assignments, the larger the crossing attribution confidence value. A crossing attribution result table is generated based on the optimal assignment, unmatched target side nodes, unmatched card-swiping side nodes, post-swiping paths, non-card-swiping continuation crossing paths, door-holding waiting paths, and closing conflict paths. The crossing attribution result table includes authorized attribution flags, post-swiping attribution flags, trailing candidate flags, object misbinding flags, door-holding waiting flags, card swipe record number, card swipe identity identifier, video target number, door line crossing sequence number, and video evidence segment number.

[0039] In this implementation scheme, by reading the passage sequence record, reverse verification record, card swipe identity sequence chain, target crossing sequence chain, and door event sequence diagram, the card swipe identity queue and the target crossing queue are organized according to the door opening interval number. A crossing attribution combination table is formed by combining the post-swipe path, the non-swipe continuation crossing path, the door-holding waiting path, and the closed conflict path, enabling constraint matching between the card swipe record number and the video target number within the same door opening interval. By constructing a card swipe crossing bipartite graph and setting time difference sub-costs, direction consistency sub-costs, post-swipe sub-costs, door-holding waiting costs, and closed conflict sub-costs, the relationship between the card swipe side node and the target side node is ensured. The assignment mechanism can take into account card swiping sequence, door crossing sequence, post-swiping, door holding waiting, and closure conflict. By assigning the target with the minimum total cost, retaining unmatched nodes, and selecting candidate assignment results with the same cost, it can cover passage scenarios where the number of card swiping records is inconsistent with the number of video targets. It generates a reliable value for passage attribution based on the optimal and suboptimal assignment results, thereby achieving the effects of attribution of swipes, identification of tailgating candidates, removal of mis-bound objects, and generation of passage attribution result tables. It effectively solves the problems in the existing technology that bind only a single card swipe record to a single door crossing target, resulting in false alarms for normal swipes, real tailgating being classified into the previous authorized passage record, and inaccurate attribution of passage objects.

[0040] Specifically, the steps for performing supplementary swipe attribution, tailgating candidate identification, and object misbinding removal based on the crossing attribution analysis results are as follows: Compare the crossing attribution confidence value and confidence threshold, where the confidence threshold is obtained from the normal distribution of historically verified crossing attribution confidence values; when the crossing attribution confidence value is greater than or equal to the confidence threshold, the crossing attribution result table is adopted, and the door opening interval corresponding to the same door opening interval number is used as the processing range for supplementary swipe attribution, tailgating candidate identification, and object misbinding removal, avoiding cross-door opening interval access to swipe records or video target numbers; records with supplementary swipe attribution markers are written into the supplementary swipe access record, ensuring that video target numbers that complete identity attribution within the same door opening interval after door opening are no longer included in the tailgating candidate record; the door opening interval number and video target number corresponding to records with tailgating candidate markers are written into the tailgating candidate record. The system records video target numbers lacking valid card swipe attribution within the same gate opening interval as candidate objects for subsequent tailgating alarm closure verification. Video target numbers with mis-binding markers are disconnected from the previous authorized access record and written into the tailgating stripping record. The previous authorized access record is limited to access records under the same gate opening interval number that have authorized attribution markers, whose gate crossing time is earlier than the gate crossing time corresponding to the video target number with mis-binding markers, and whose gate crossing sequence numbers are adjacent. The tailgating stripping record carries the previous authorized access record number, the candidate object record number after stripping, and the video evidence segment number. When the crossing attribution confidence value is less than the confidence threshold, updating the supplementary card swipe access record and the tailgating stripping record is paused. The corresponding video target number is written into the attribution pending verification record, and the crossing attribution combination table and the video evidence segment number are retained for evidence verification.

[0041] In this implementation scheme, by comparing the reliable value of the crossing attribution with the reliable threshold, the crossing attribution result table is adopted under the reliable attribution condition. The supplementary attribution mark, the tailing candidate mark, and the object misbinding mark are converted into supplementary passage record, tailing candidate record, and tailing stripping record, respectively. This allows normal supplementary targets within the same door opening interval to be excluded from the tailing candidate, video target numbers lacking valid card swipe attribution to be included in the subsequent tailing alarm closure verification, and video target numbers that are incorrectly attached to the previous authorized passage record to be stripped out. By limiting the previous authorized passage record to passage records with authorized attribution marks under the same door opening interval number, whose door line crossing time is earlier than the object misbinding target, and whose door line crossing sequence number is adjacent, cross-interval or cross-sequence mis-stripping is avoided. When the crossing attribution reliable value is insufficient, an attribution pending verification record is written, thereby improving the accuracy of supplementary attribution, tailing candidate identification, and object misbinding stripping.

[0042] Specifically, the steps for tailgating alarm closure verification, combining the crossing attribution result table, tailgating candidate records, supplementary passage records, tailgating stripping records, and attribution pending verification records, are as follows: Read the crossing attribution result table, supplementary passage records, tailgating candidate records, tailgating stripping records, attribution pending verification records, and video evidence segment numbers; form a candidate object row index using the door opening interval numbers in the tailgating candidate records and tailgating stripping records and the video target number; generate door line crossing evidence bits based on the door line crossing sequence number and video evidence segment number in the tailgating candidate records; and determine the evidence bits based on whether the authorized attribution flag or the attribution flag is hit in the crossing attribution result table. The system generates a card swipe gap evidence bit when the swipe record with the attribution mark is supplemented. When no record in the supplemented passage record matches the video target number of the trailing candidate record, a supplemented swipe exclusion evidence bit is generated. An object stripping evidence bit is generated based on the previous authorized passage record number, the stripped candidate record number, and the video evidence segment number in the trailing stripping record. The previous authorized passage record number is the passage record number within the same door opening interval that, after being sorted by door line crossing time, is located before the stripped candidate record number and hits the authorized attribution mark. When there is a record in the attribution pending verification that matches the video target number of the candidate row... During recording, a temporary suppression bit for evidence collection is generated; an evidence column index is formed using the door line crossing evidence bit, card swipe gap evidence bit, re-swipe exclusion evidence bit, object stripping evidence bit, and temporary suppression bit for evidence collection, constructing a trailing evidence closure matrix; according to the candidate object row index and the evidence column index, the matrix elements that hit each evidence bit are assigned a value of one, and the matrix elements that do not hit are assigned a value of zero; a trailing closure weight vector is generated according to the evidence column index, and the weights of the door line crossing evidence bit, card swipe gap evidence bit, re-swipe exclusion evidence bit, and object stripping evidence bit in the trailing closure weight vector are positive weights, and the positive weights are based on historical samples. The contribution ratio of each evidence bit to the tailgating sample marker is obtained and normalized to the sum of four positive weights of one. The evidence collection delay suppression bit corresponds to a negative weight. The negative weight is used to deduct the impact of the pending record on the tailgating alarm closure, and the absolute value of the negative weight is not greater than one. The tailgating evidence closure matrix and the tailgating closure weight vector are multiplied to obtain the candidate object closure vector. The vector components corresponding to the candidate object row are truncated according to the value range of zero to one and used as the tailgating alarm closure value. The larger the tailgating alarm closure value, the more the video target number corresponding to the candidate object row meets the conditions for generating the tailgating warning report.

[0043] As shown in Table 1, this is the tail alarm closure value verification data table based on the candidate object evidence bit hit result in this embodiment. In the table, QJ001, QJ002 and QJ003 are example codes for the door opening interval number, and VT01 to VT05 are example codes for the video target number. The first candidate object has a door opening section number of QJ001, a video target number of VT01, a door line crossing evidence bit of 1, a card swipe gap evidence bit of 0, a supplementary swipe exclusion evidence bit of 0, an object stripping evidence bit of 0, and an evidence collection delay suppression bit of 0. The calculated tailing alarm closure value is 0.22. The second candidate object has a door opening section number of QJ001, a video target number of VT02, a door line crossing evidence bit of 1, a card swipe gap evidence bit of 1, a supplementary swipe exclusion evidence bit of 1, an object stripping evidence bit of 1, and an evidence collection delay suppression bit of 0. The calculated tailing alarm closure value is 1.00. The third candidate object has a door opening section number of QJ002, a video target number of VT03, a door line crossing evidence bit of 1, a card swipe gap evidence bit of 1, and a card swipe gap evidence bit of 0. The following candidate object has a door opening interval number of QJ003, a video target number of VT04, a door line crossing evidence bit of 1, a card swipe gap evidence bit of 1, a card swipe exclusion evidence bit of 0, an object stripping evidence bit of 0, and an evidence collection suspension suppression bit of 0. The calculated tailing alarm closure value is 0.76. The following candidate object has a door opening interval number of QJ003, a video target number of VT04, a door line crossing evidence bit of 1, a card swipe gap evidence bit of 1, a card swipe exclusion evidence bit of 0, an object stripping evidence bit of 0, and an evidence collection suspension suppression bit of 1. The calculated tailing alarm closure value is 0.15. The following candidate object has a door opening interval number of QJ003, a video target number of VT05, a door line crossing evidence bit of 1, a card swipe gap evidence bit of 0, a card swipe exclusion evidence bit of 1, an object stripping evidence bit of 0, and an evidence collection suspension suppression bit of 0. The calculated tailing alarm closure value is 0.46.

[0044] Table 1. Tail Alarm Closure Value Verification Data Table Based on Candidate Object Evidence Bit Hit Results

[0045]

[0046] like Figure 4The diagram shows a trailing evidence closure matrix based on the candidate object row index and evidence column index in this embodiment. The candidate object row index is formed by the door opening interval number and the video target number. The evidence column index is formed by the door line crossing evidence bit, card swipe gap evidence bit, supplementary swipe exclusion evidence bit, object stripping evidence bit, and evidence collection postponement / suppression bit. The matrix elements' 1 and 0 indicate whether the corresponding evidence bit is hit. Table 1 shows the trailing alarm closure value verification data table based on the candidate object evidence bit hit results in this embodiment. The table records the evidence for each video target number under the corresponding door opening interval number. The system records the hit status, tailing alarm closure value, and judgment results. Among them, video target number VT02 simultaneously hits the door line crossing evidence bit, card swipe gap evidence bit, supplementary swipe exclusion evidence bit, and object stripping evidence bit, and the tailing alarm closure value reaches the warning condition, generating a tailing warning report. Video target number VT04 hits the evidence collection suspension suppression bit. Although it has the door line crossing evidence bit and card swipe gap evidence bit, the tailing alarm closure value is insufficient to generate a tailing warning report, and it is written into the tailing closure insufficiency retention record. This demonstrates the tailing alarm closure verification process for diverting strong tailing objects, weak closure objects, and objects whose evidence collection is suspended.

[0047] In this implementation plan, by uniformly mapping the crossing attribution result table, re-swiping passage records, tailgating candidate records, tailgating stripping records, and attribution pending verification records to the candidate object row index and evidence column index, the door line crossing evidence bits, card swipe gap evidence bits, re-swiping exclusion evidence bits, object stripping evidence bits, and evidence collection postponement suppression bits can form a computable tailgating evidence closure matrix. At the same time, positive weights are used to strengthen the supporting role of door line crossing, card swipe gap, re-swiping exclusion, and object stripping in tailgating judgment, while negative weights are used to deduct the impact of attribution pending verification records on alarm closure. This allows the tailgating alarm closure value to reflect the tailgating risk level of the video target number corresponding to the candidate object on a unified scale, thereby providing a quantitative basis for tailgating early warning report generation, passage tracing record output, and tailgating closure insufficiency retention.

[0048] Specifically, the steps for generating a tailgating warning report, passage tracing record, and tailgating closure insufficiency retention record based on the tailgating alarm closure verification result are as follows: The tailgating alarm closure value and closure threshold are compared, where the closure threshold is obtained from the distribution of tailgating alarm closure values ​​corresponding to historical tailgating and non-tailgating passage samples; when the tailgating alarm closure value is greater than or equal to the closure threshold, a tailgating warning report is generated, the video target number is marked as a suspected tailgating object, and within the same gate opening interval number, records whose gate crossing time is earlier than the gate crossing time corresponding to the suspected tailgating object and have already hit the authorization attribution mark are used as authorized passage candidate records, and the previous authorized passage is determined by sorting the gate crossing times from nearest to farthest. Record number; when authorized access candidate records have the same gate crossing time, sort them in ascending order of the absolute value of the difference between the card swipe time and the gate crossing time corresponding to the suspected tailgating object to determine the previous authorized access record number; write the previous authorized access record number, the candidate object record number after stripping, and the video evidence segment number into the access tracing record, and write the gate crossing record, card swipe gap record, supplementary swipe exclusion record, and object stripping record associated with the evidence bit into the monitoring log; when the tailgating alarm closure value is less than the closure threshold, write the corresponding video target number into the tailgating closure insufficiency retention record, retain the tailgating evidence closure matrix, crossing attribution combination table, and video evidence segment number, and do not generate a tailgating warning report for the time being.

[0049] In this implementation plan, a warning judgment is performed on candidate objects by comparing the tailgating alarm closure value and the closure threshold. When the tailgating alarm closure value is greater than or equal to the closure threshold, the previous authorized passage record number is determined within the same gate opening interval number based on the gate crossing time and card swiping time, and a tailgating warning report, passage tracing record and monitoring log are generated, thereby achieving the effects of tailgating object location, previous authorized passage tracing and evidence retention. When the tailgating alarm closure value is less than the closure threshold, a tailgating closure insufficiency retention record is generated and the tailgating evidence closure matrix, crossing attribution combination table and video evidence segment number are retained, effectively solving the problem of high false alarm risk in scenarios where the tailgating warning object is not unique and the evidence is insufficient in the existing technology.

[0050] Specifically, this embodiment provides an abnormal event identification and early warning system based on campus multi-source monitoring data, applied to an abnormal event identification and early warning method based on campus multi-source monitoring data. It includes a data acquisition and preprocessing module for acquiring campus multi-source monitoring data and performing preprocessing operations on the data. Specifically targeting campus access control scenarios, it acquires access-related data from the processes of card swiping, door opening and closing, door area video, and door line detection. It performs time calibration, duplicate card swiping merging, failed card swiping isolation and archiving, enabling interval slicing, door line mapping, trajectory interception, and crossing positioning on data from different sources, thus establishing a card swiping identity sequence chain, a target crossing sequence chain, a door area access mapping table, and a door area access detection system. The chain can form a unified data foundation around the numbering of the same door opening interval; the door opening and holding reversal discrimination module is used to read the door area passage detection chain and the historical passage sample map library, perform passage reversal analysis, and generate passage sequence records and reversal verification records based on the passage reversal analysis results. Specifically, it organizes the card swiping time, door opening event time, door pushing action time, target appearance time, door line crossing time, and door closing event time in sequence through the door event sequence diagram, and combines it with the historical passage sample map library to identify the follow-up path after card swiping, the passage path without card swiping, the door holding waiting path, and the closing conflict path, thereby distinguishing between card swiping first passage and reversal passage that requires further verification; the follow-up swiping attribution and tailgating The image stripping module is used to perform crossing attribution analysis by calling the reverse verification record, card swipe identity sequence chain, and target crossing sequence chain. Based on the crossing attribution analysis results, it performs supplementary swipe attribution, tailgating candidate identification, and object misbinding stripping. Specifically, it uses the door opening interval number as the processing range, combining the interval card swipe identity queue, interval target crossing queue, and crossing attribution combination table to generate a crossing attribution result table. Based on the supplementary swipe attribution marker, tailgating candidate marker, and object misbinding marker, it forms supplementary passage records, tailgating candidate records, and tailgating stripping records respectively, enabling the differentiation and processing of normal supplementary swipe targets, video target numbers lacking valid card swipe attribution, and video target numbers incorrectly linked to the previous authorized passage record; tailgating certificate. According to the closing and early warning output module, it is used to combine the crossing attribution result table, tailgating candidate records, supplementary passage records, tailgating stripping records, and attribution pending verification records to perform tailgating alarm closing verification. Based on the tailgating alarm closing verification results, it generates tailgating early warning reports, passage tracing records, and tailgating closure insufficiency retention records. Specifically, it constructs a tailgating evidence closing matrix using the candidate object row index and evidence column index, and combines the door line crossing evidence bit, card swipe gap evidence bit, supplementary swipe exclusion evidence bit, object stripping evidence bit, and evidence collection suspension and suppression bit to form the tailgating alarm closing basis, so that the suspected tailgating object can form a traceable association with the video evidence segment number, the previous authorized passage record number, and the candidate object record number after stripping.

[0051] In this implementation plan, a gate access detection chain is formed through a data acquisition and preprocessing module. Then, a gate opening and holding reversal discrimination module identifies the post-swiping path, the no-card-swipe continuation crossing path, the door-holding waiting path, and the closing conflict path. The swipe attribution and tailgating object stripping module processes the swipe attribution mark, tailgating candidate mark, and object misbinding mark separately. Finally, the tailgating evidence closure and early warning output module combines the crossing attribution result table, tailgating candidate record, swipe passage record, tailgating stripping record, and attribution pending verification record to complete the tailgating alarm closure verification. This enables the campus multi-source monitoring data to form a complete processing chain from acquisition, reversal discrimination, crossing attribution to evidence closure, thereby improving the accuracy of distinguishing normal swipe targets, suspected tailgating objects, and object misbinding targets, and providing traceable data basis for tailgating early warning reports, passage tracing records, and tailgating closure insufficiency retention records.

[0052] It should be noted that, in this document, the terms "comprising," "including," and any other variations 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 a process, method, article, or apparatus. Specific examples have been used in this document to illustrate the principles and implementation methods of the present invention. These examples are merely for the purpose of helping to understand the method and core ideas of the present invention. The above descriptions are only preferred embodiments of the present invention. It should be pointed out that, due to the limitations of written expression and the objective existence of infinite specific structures, those skilled in the art can make several improvements, modifications, or variations without departing from the principles of the present invention, and can also combine the above technical features in an appropriate manner. These improvements, modifications, variations, or combinations, or the direct application of the concept and technical solution of the present invention to other situations without modification, should all be considered within the scope of protection of the present invention.

Claims

1. An abnormal event identification and early warning method based on multi-source campus monitoring data, characterized in that, Includes the following steps: S1 collects multi-source monitoring data from the campus and performs preprocessing operations on the multi-source monitoring data from the campus. S2, read the gate area access detection chain and historical access sample map library, perform access reversal analysis, and generate access sequence passing records and reversal verification records based on the access reversal analysis results; S3 calls the reverse verification record, the card swipe identity sequence chain and the target crossing sequence chain to perform crossing attribution analysis, and performs supplementary swipe attribution, tailing candidate identification and object misbinding stripping according to the crossing attribution analysis results; S4 combines the crossing attribution result table, tailing candidate records, supplementary passage records, tailing stripping records, and attribution pending verification records to perform tailing alarm closure verification. Based on the tailing alarm closure verification results, a tailing warning report, passage tracing records, and tailing closure insufficiency retention records are generated.

2. The abnormal event identification and early warning method based on campus multi-source monitoring data according to claim 1, characterized in that: The specific steps for collecting multi-source monitoring data on campus are as follows: Collect multi-source monitoring data on campus: Obtain access control point number, card swipe record number, card swipe identity identifier, card swipe time, card swipe result code, and passage record number through the card swipe log interface of the access control controller; obtain door number, door opening event time, door closing event time, door opening duration, and door event code through the opening and closing event logs of door magnetic sensors and door lock controllers; obtain video channel number, camera number, door line number, video target number, target appearance time, door pushing action time, door line crossing time, crossing direction mark, door holding waiting mark, and video evidence segment number through the video stream of the door area camera and the door line detection program; obtain access control point number, door number, camera number, door line number, door line position, and door passage direction through the access control point configuration table; obtain normal re-swipe sample mark, tailgating sample mark, door holding waiting sample mark, sample card swipe trigger point, sample door opening point, sample door pushing action point, sample target appearance point, sample door line crossing point, and sample door closing point through historical passage samples. 3.The method of claim 1, wherein the method further comprises: The specific steps for preprocessing multi-source monitoring data on campus are as follows: The system performs time calibration, duplicate swipe merging, and failed swipe isolation archiving on access control point numbers, card swipe record numbers, card swipe identification, card swipe time, and card swipe result codes, generating a card swipe identity sequence chain based on the card swipe time. It also performs opening interval slicing on door numbers, door opening event times, door closing event times, door opening duration, and door event codes, generating door opening interval numbers and door opening intervals. Furthermore, it performs door line mapping, trajectory interception, and crossing positioning on video channel numbers, camera numbers, door line numbers, video target numbers, target appearance times, door pushing action times, door line crossing times, crossing direction markers, door holding waiting markers, and video evidence segment numbers, generating a target crossing sequence chain. Finally, it performs point mapping on access control point numbers, door numbers, camera numbers, and door line numbers, generating a door area access mapping table. Using the door opening interval number as the attachment index, it writes the card swipe identity sequence chain, target crossing sequence chain, door area access mapping table, and historical access sample data into the door area access detection chain. 4.The method of claim 1, wherein the method further comprises: The specific steps for reading the gate area access detection chain and historical access sample map library to perform access reversal analysis are as follows: Read the door zone access detection chain, using the door opening interval number as an index, and combine the card swipe time, door opening event time, door pushing action time, target appearance time, door line crossing time, and door closing event time to construct a door event sequence graph. The door event sequence graph includes a set of graph vertices and a set of directed edges. The set of graph vertices includes the card swipe trigger point, door opening point, door pushing action point, target appearance point, door line crossing point, and door closing point. The set of directed edges includes directed edges indicating the sequence of events and directed edges indicating the door opening interval affiliation. The system reads the historical passage sample graph library, which contains sample graphs consisting of sample card swipe trigger points, sample door opening points, sample door pushing action points, sample target appearance points, sample door line crossing points, sample door closing points, and directed edges representing the sequence of sample events. Using graph sequence dynamic regularization technology, the system aligns the time sequence of the door event sequence graph with the sample graphs in the historical passage sample graph library and matches the directed edge paths. Based on the differences in the time sequence of the graph points and the differences in the directed edge paths, it identifies follow-up swipe paths, non-swipe continuation crossing paths, door holding waiting paths, and closing conflict paths, and generates passage reversal values.

5. The method of claim 1, wherein the method further comprises: The specific steps for generating the passage sequence record and reversal verification record based on the passage reversal analysis results are as follows: Compare the passage reversal value with the reversal threshold. When the passage reversal value is less than the reversal threshold, bind the card swipe record and the door line crossing record according to the card swipe first rule to generate a passage sequence record. When the passage reversal value is greater than or equal to the reversal threshold, the one-to-one binding relationship between a single card swipe record and a single door target is temporarily not accepted. Instead, the card swipe identity sequence chain, target passage sequence chain, door event sequence diagram, post-swipe path, non-swipe continuation passage path, door holding waiting path, and closing conflict path are written into the reversal verification record. 6.The method of identifying and warning abnormal events based on campus multi-source monitoring data according to claim 1, characterized in that: The specific steps for performing the crossing attribution analysis by calling the reverse verification record, the card swipe identity sequence chain, and the target crossing sequence chain are as follows: The system reads the passage sequence record, reverse verification record, card swipe identity sequence chain, target crossing sequence chain, and door event sequence diagram. Using the door opening interval number as an index, it arranges the card swipe record numbers by card swipe time to generate an interval card swipe identity queue, and arranges the video target numbers by door line crossing time to generate an interval target crossing queue. It also reads the door line crossing sequence number and video evidence segment number from the target crossing sequence chain. From the reverse verification record, it reads the follow-up path after card swipe, the non-card swipe continuation crossing path, the door-holding waiting path, and the closing conflict path. It then writes the interval card swipe identity queue, the interval target crossing queue, the follow-up path after card swipe, the non-card swipe continuation crossing path, the door-holding waiting path, and the closing conflict path into the crossing attribution combination table. Finally, it uses the interval card swipe identity... The queue serves as the card-swiping side node, and the interval target crossing queue serves as the target side node, constructing a card-swiping crossing bipartite graph. The assignment cost between the card-swiping side node and the target side node is generated based on the time difference between the card-swiping time and the door-line crossing time, the crossing direction marker, the re-swiping post-path, the door-holding waiting path, and the closed conflict path. With the goal of minimizing the total assignment cost, a card-swiping crossing assignment model is constructed by combining the crossing assignment combination table and the door event sequence graph, generating a crossing assignment confidence value and a crossing assignment result table. The crossing assignment result table includes authorized assignment markers, re-swiping assignment markers, trailing candidate markers, object misbinding markers, door-holding waiting markers, card-swiping record number, card-swiping identity identifier, video target number, door-line crossing sequence number, and video evidence segment number.

7. The abnormal event identification and early warning method based on campus multi-source monitoring data according to claim 1, characterized in that: The specific steps for performing attribution correction, tailing candidate identification, and object mis-binding removal based on the crossing attribution analysis results are as follows: Compare the crossing attribution confidence value and confidence threshold. If the crossing attribution confidence value is greater than or equal to the confidence threshold, the crossing attribution result table is adopted. Records with supplementary attribution markers are written into the supplementary passage record. The door opening interval number and video target number corresponding to the records with tailing candidate markers are written into the tailing candidate record. The video target number with object misbinding markers is disconnected from the previous authorized passage record and written into the tailing stripping record. The tailing stripping record carries the previous authorized passage record number, the stripped candidate object record number, and the video evidence segment number. When the crossing attribution confidence value is less than the confidence threshold, the update of the supplementary passage record and the tail stripping record is paused. The corresponding video target number is written into the attribution pending verification record, and the crossing attribution combination table and video evidence segment number are retained for evidence verification. 8.The method of claim 1, wherein the method further comprises: The specific steps for performing tailgating alarm closure verification by combining the crossing attribution result table, tailgating candidate records, supplementary passage records, tailgating stripping records, and attribution pending verification records are as follows: Read the crossing attribution result table, re-swipe passage records, tailgating candidate records, tailgating stripping records, attribution pending verification records, and video evidence segment numbers. Form a candidate object row index using the door opening interval number in the tailgating candidate records and tailgating stripping records and the video target number. Generate a door line crossing evidence bit based on the door line crossing sequence number and video evidence segment number in the tailgating candidate records. Generate a card swipe gap evidence bit based on records in the crossing attribution result table that neither match the authorized attribution marker nor the re-swipe attribution marker. Generate a re-swipe exclusion evidence bit when there is no record in the re-swipe passage record that matches the video target number of the tailgating candidate record. Generate a re-swipe exclusion evidence bit based on the previous... An object stripping evidence bit is generated from an authorized access record number, a candidate object record number after stripping, and a video evidence segment number. When there is a record in the pending record that matches the video target number of the candidate object row, an evidence collection suspension bit is generated. An evidence column index is formed using the door line crossing evidence bit, card swipe gap evidence bit, re-swipe exclusion evidence bit, object stripping evidence bit, and evidence collection suspension bit, and a trailing evidence closed matrix is ​​constructed. According to the candidate object row index and the evidence column index, the matrix elements that hit each evidence bit are assigned a value of one, and the matrix elements that do not hit are assigned a value of zero. A trailing closed weight vector is generated according to the evidence column index, where the evidence collection suspension bit corresponds to a negative weight. Perform matrix multiplication between the trailing evidence closure matrix and the trailing closure weight vector to obtain the candidate object closure vector, and use the vector components of the candidate object row as the trailing alarm closure value. 9.The method of identifying and warning abnormal events based on campus multi-source monitoring data according to claim 1, characterized in that: The specific steps for generating a tailgating early warning report, a passage tracing record, and a tailgating insufficiency retention record based on the tailgating alarm closure verification result are as follows: Compare the tailgating alarm closure value and closure threshold. When the tailgating alarm closure value is greater than or equal to the closure threshold, a tailgating warning report is generated. The video target number is marked as a suspected tailgating object. The previous authorized passage record number, the candidate object record number after stripping, and the video evidence segment number are written into the passage tracing record. The door line crossing record, card swipe gap record, supplementary swipe exclusion record, and object stripping record associated with the evidence bit are written into the monitoring log. When the tailing alarm closure value is less than the closure threshold, the corresponding video target number is written into the tailing closure insufficiency retention record, and the tailing evidence closure matrix, crossing attribution combination table and video evidence segment number are retained. The tailing warning report is not generated at this time.

10. An abnormal event identification and early warning system based on campus multi-source monitoring data, applied to the abnormal event identification and early warning method based on campus multi-source monitoring data as described in any one of claims 1-9, characterized in that, include: The data acquisition and preprocessing module is used to acquire multi-source monitoring data on campus and perform preprocessing operations on the multi-source monitoring data on campus. The gate opening and holding reversal discrimination module is used to read the gate area access detection chain and historical access sample map library, perform access reversal analysis, and generate access sequence passage records and reversal verification records based on the access reversal analysis results; The module for re-swiping attribution and tailing object stripping is used to call the reverse verification record, the card swipe identity sequence chain and the target crossing sequence chain to perform crossing attribution analysis, and to perform re-swiping attribution, tailing candidate identification and object misbinding stripping based on the crossing attribution analysis results. The tailing evidence closure and early warning output module is used to combine the crossing attribution result table, tailing candidate records, supplementary passage records, tailing stripping records, and attribution pending verification records to perform tailing alarm closure verification. Based on the tailing alarm closure verification results, it generates tailing early warning reports, passage tracing records, and tailing closure insufficiency retention records.

Images