Campus ai security access control linkage control and early warning method and system

Abstract

The application relates to the technical field of campus security and protection, in particular to a campus AI security and protection door access linkage control and early warning method and system, which extracts time sequence position point sets and identity information through collection of video monitoring and door access authentication data, constructs an identity trajectory double-layer mapping network and identifies a conflict structure, marks corresponding personnel as to-be-verified objects, extracts current and historical trajectory units of the to-be-verified objects, calculates time distribution of the to-be-verified objects to reach each door access device according to the historical trajectory units, establishes a door access cooperative control matrix with rows corresponding to door access devices, columns corresponding to time windows and elements representing control states, fills in a limit state in a corresponding time window, issues a time sequence control instruction and pushes an early warning message, and collects execution feedback and confirmation results. The application can realize rapid identification of abnormal personnel and door access linkage control and management, effectively improves the intelligent level of campus security and protection and the emergency response efficiency.

Description

Technical Field

[0001] This invention relates to the field of campus security technology, and in particular to a campus AI security access control and early warning method and system. Background Technology

[0002] In the field of campus security, video surveillance systems and access control systems are typically deployed and operated as independent subsystems. Video surveillance systems are responsible for real-time image capture and recording of public areas on campus, relying on manual review by security personnel or post-event playback to detect abnormal behavior. Access control systems, on the other hand, use methods such as card swiping, fingerprint, or facial recognition to verify the identity of individuals entering or exiting specific buildings or areas, and generate entry and exit logs. The two systems generally cooperate only through simple linkage rules; for example, when an access control system triggers an unauthorized entry alarm, nearby surveillance footage is retrieved for manual verification, or the access control record and video playback time are roughly aligned for later traceability.

[0003] Video surveillance data and access control authentication data are stored on separate platforms, lacking a unified data fusion and processing mechanism. This makes it difficult to accurately correlate personnel location and identity information over time, hindering security personnel from promptly determining the true connection between individuals appearing on surveillance footage and their access control authentication records. This data fragmentation means that cross-regional personnel movement trajectories can only be reconstructed through manual comparison of multiple playback videos and access control logs, resulting in low efficiency and a high risk of missing crucial information.

[0004] Furthermore, existing linkage control methods mostly rely on static preset rules, such as setting prohibition or alarm policies only for specific time periods and specific access control points, and cannot dynamically adjust control measures based on the historical activity patterns of personnel. For personnel who have exhibited abnormal behavior or pose security risks, the system cannot proactively predict their possible movement paths, let alone implement coordinated restrictions on relevant access control equipment in advance. It can often only respond passively after a dangerous event occurs, thus limiting both early warning capabilities and prevention effectiveness. Summary of the Invention

[0005] This invention provides a campus AI security access control and early warning method and system, which can solve the problems in the prior art.

[0006] A first aspect of this invention provides a campus AI security access control and early warning method, comprising:

[0007] Collect video surveillance data and access control authentication data within the campus; extract the temporal location set of personnel from the video surveillance data; and extract identity information and authentication time from the access control authentication data.

[0008] A two-layer identity trajectory mapping network is constructed. The trajectory unit is formed by clustering time-series location points as the base layer, and the association edge is established between identity information and trajectory unit as the association layer. The weight of the association edge is determined by the temporal proximity between the authentication time and the location time. Connectivity analysis is performed on the two-layer mapping network to identify conflict structures.

[0009] Mark the personnel corresponding to the conflict structure as objects to be verified, and extract the current location and historical trajectory units of the objects to be verified.

[0010] Establish an access control collaborative control matrix. The rows of the matrix correspond to access control devices, the columns correspond to time windows, and the elements represent control states. Calculate the time distribution of the object to be verified arriving at each access control device based on historical trajectory units, and fill the corresponding time windows of the matrix with restriction states.

[0011] Based on the access control collaborative control matrix, timing control commands are sent to the access control devices and early warning messages are pushed to the security terminal. The execution feedback from the access control devices and the confirmation results from the security terminal are collected.

[0012] A two-layer identity trajectory mapping network is constructed. The trajectory units, formed by clustering temporal location points, serve as the base layer. The association layers are established by linking identity information with these trajectory units. The weights of the association edges are determined by the temporal proximity between the authentication time and the location point time. Connectivity analysis is performed on the two-layer mapping network to identify conflict structures, including:

[0013] Incremental clustering is performed on the temporal location point set. When a location point is received, the spatiotemporal distance to the end of the existing trajectory unit is calculated. When the spatiotemporal distance meets the continuation condition, the location point is added to the existing trajectory unit. When the spatiotemporal distance does not meet the continuation condition, a new trajectory unit is created.

[0014] For the authentication time of identity information, the tolerance range is expanded on the time axis, and the trajectory unit is searched within the tolerance range. The association strength is determined according to the distance from the authentication time to the time boundary of the trajectory unit, and an association edge carrying the association strength is established between the identity information and the trajectory unit.

[0015] Perform bidirectional propagation operation on the two-layer mapping network, propagate identity labels from identity nodes to trajectory nodes, and return verification results from trajectory nodes to identity nodes. When identity labels conflict or verification results fail, set conflict markers on the associated edges.

[0016] Scan the associated edges carrying conflict markers, trace the identity nodes and trajectory nodes corresponding to the conflict markers to construct a local subnetwork, and count the out-degree of identity nodes and the in-degree of trajectory nodes in the local subnetwork.

[0017] When the in-degree of a trajectory node is multiple, it is identified as an identity competition conflict; when the out-degree of an identity node is multiple, it is identified as a trajectory splitting conflict. The conflict type is then output along with the local sub-network.

[0018] A bidirectional propagation operation is performed on the two-layer mapping network, propagating identity labels from identity nodes to trajectory nodes and returning verification results from trajectory nodes to identity nodes. When identity labels conflict or verification results fail, conflict markers are set on the associated edges, including:

[0019] Assign a propagation timestamp to each associated edge in the two-layer mapping network. The propagation timestamp identifies the propagation order of the identity label along the associated edge and assigns values ​​to the propagation timestamps according to the association strength of the associated edge.

[0020] Activate associated edges sequentially according to the order of propagation timestamps, extract identity tags from the identity nodes connected by the associated edges and propagate them to the trajectory nodes, record the binding relationship between the propagation timestamp and the identity tag when the trajectory node receives the identity tag, and store it in the trajectory node's receiving record;

[0021] For trajectory nodes with multiple binding relationships in the received records, extract the identity tags corresponding to the multiple binding relationships, calculate the attribute similarity between the identity tags, determine the tag conflict when the attribute similarity is lower than the similarity threshold, and determine the tag consistency when the attribute similarity is not lower than the similarity threshold. The determination results of tag conflict or tag consistency are then encapsulated as verification results.

[0022] The verification results are sent back to the identity node along the associated edge in reverse order of the propagation timestamp. When the identity node receives the verification results, it parses the judgment results. When the judgment results show that the labels are consistent, the verification process is completed. When the judgment results show that the labels conflict, a conflict marker is set on the corresponding associated edge.

[0023] The personnel corresponding to the conflict structure are marked as objects to be verified, and the current position and historical trajectory units of the objects to be verified are extracted, including:

[0024] Extract identity nodes from the conflict structure, read the identity information carried by the identity nodes to determine the corresponding personnel, assign verification priority identifiers to the personnel and mark them as objects to be verified;

[0025] Tracing back the trajectory nodes associated with the object to be verified from the conflict structure, extracting the trajectory units corresponding to the trajectory nodes to construct a trajectory set, and arranging the trajectory units in the trajectory set in time sequence according to the time boundary;

[0026] In the temporal arrangement, identify trajectory units whose time boundaries are later than the current time, extract the end position point as the current position of the object to be verified, and identify trajectory units whose time boundaries are earlier than the current time in the temporal arrangement as historical trajectory units;

[0027] For each historical trajectory unit, a set of temporal location points is extracted. Spatial clustering and temporal continuity analyses are performed on the set of temporal location points. Based on the spatial clustering analysis, the activity area type of the historical trajectory unit is labeled, and based on the temporal continuity analysis, the behavior pattern type of the historical trajectory unit is labeled.

[0028] The current location, historical trajectory unit, activity area type, and behavior pattern type of the object to be verified are associated and stored to generate a spatiotemporal behavior profile.

[0029] Establish an access control collaborative control matrix. Rows in the matrix correspond to access control devices, columns correspond to time windows, and elements represent control states. Calculate the time distribution of the object to be verified arriving at each access control device based on historical trajectory units, and fill the corresponding time windows in the matrix with restriction states, including:

[0030] Read the spatial coordinates and device number of the access control device, and assign row indexes to the access control collaborative control matrix according to the device number; set the prediction period and time window interval, divide the time window sequence and assign column indexes to generate the access control collaborative control matrix;

[0031] Extract the temporal position data of the historical trajectory units of the object to be verified, perform trajectory morphology analysis to extract path curvature and direction change rate, and identify movement habit characteristics based on path curvature and direction change rate;

[0032] By coupling mobile habit features with the current location, the position evolution process of the object to be verified is deduced within the prediction period. Evolution nodes and evolution times are extracted from the position evolution process and combined to generate a spatiotemporal evolution sequence.

[0033] Determine the spatial reachability from evolution nodes to access control devices in the spatiotemporal evolution sequence, identify associated access control devices based on spatial reachability, and establish a binding relationship between the device number of the associated access control devices and the evolution time.

[0034] Based on the binding relationship, locate the row index corresponding to the device number in the access control collaborative control matrix and the column index of the time window to which the evolution time belongs, fill the restriction status at the location position and attach the verification priority identifier of the object to be verified.

[0035] Extract the temporal position data of the historical trajectory units of the object to be verified, perform trajectory morphology analysis to extract path curvature and direction change rate, and identify movement habit features based on path curvature and direction change rate, including:

[0036] Extract temporal position data from the historical trajectory units of the object to be verified, convert the temporal position data into a spatial coordinate point sequence, and segment the spatial coordinate point sequence according to temporal continuity to form continuous trajectory segments;

[0037] A reference vector is constructed by the start and end points of a continuous trajectory segment. The vertical distance from each point on the continuous trajectory segment to the reference vector is extracted. The vertical distance is statistically processed to generate the path curvature.

[0038] Extract the movement direction of the position points on the continuous trajectory segment and convert it into an angle representation. Calculate the change in adjacent angles to construct an angle change sequence. Perform wave analysis on the angle change sequence to extract the wave amplitude as the direction change rate.

[0039] Establish a mapping table between path curvature and direction change rate. The mapping table is constructed with path curvature as the row index and direction change rate as the column index to form a two-dimensional structure. The index intersection position stores the type identifier of the movement habit feature.

[0040] Based on the path curvature and direction change rate of the object to be verified, the corresponding position is retrieved from the mapping table, and the type identifier stored at the corresponding position is read to determine the movement habit characteristics.

[0041] Based on the access control collaborative control matrix, timing control commands are issued to the access control devices and early warning messages are pushed to the security terminal. The execution feedback from the access control devices and the confirmation results from the security terminal are collected, including:

[0042] Calculate the spatial distance between each access control device based on the current location of the object to be verified, mark the access control devices that are closer as high priority and generate a timing control instruction queue;

[0043] Control commands are sequentially retrieved from the command queue, each command is assigned a unique identifier and effective time and written into the message header, and then pushed to the access control device through the message bus;

[0044] A multi-layered early warning message containing identity information, trajectory coordinate sequence, and restricted status location distribution is constructed, encapsulated, and pushed to the security terminal. The security terminal unfolds the multi-layered message according to the hierarchy, triggering identity verification, trajectory backtracking, and policy review processes respectively. The results of each process are collected and processed independently and then merged into a confirmation result.

[0045] After receiving the instruction, the access control device establishes an execution trace based on the instruction identifier. At the effective time, it starts control and monitors the authentication event, binds the authentication event and control response into an event response pair and sends it back as execution feedback. It cross-verifies the confirmation result with the event response pair. When the processing result is inconsistent with the response status, a conflict alarm is generated and fed back to the access control collaborative control matrix for policy correction.

[0046] A second aspect of this invention provides a campus AI security access control and early warning system, comprising:

[0047] The data acquisition unit is used to collect video surveillance data and access control authentication data on campus, extract the time-series location points of personnel from the video surveillance data, and extract identity information and authentication time from the access control authentication data.

[0048] The mapping construction unit is used to construct a two-layer identity trajectory mapping network. The trajectory unit is formed by clustering time-series location points as the base layer, and the association edge is established between identity information and trajectory unit as the association layer. The weight of the association edge is determined by the temporal proximity between the authentication time and the location point time. The connectivity analysis of the two-layer mapping network is performed to identify conflict structures.

[0049] The object recognition unit is used to mark the personnel corresponding to the conflict structure as objects to be verified, and to extract the current position and historical trajectory of the objects to be verified.

[0050] The matrix establishment unit is used to establish the access control collaborative control matrix. The rows of the matrix correspond to the access control devices, the columns correspond to the time windows, and the elements represent the control status. Based on the historical trajectory unit, the time distribution of the object to be verified arriving at each access control device is calculated, and the restriction status is filled in the corresponding time window of the matrix.

[0051] The control execution unit is used to issue timing control commands to the access control equipment and push early warning messages to the security terminal according to the access control collaborative control matrix, and to collect the execution feedback of the access control equipment and the confirmation results of the security terminal.

[0052] A third aspect of the present invention provides an electronic device, comprising:

[0053] processor;

[0054] Memory used to store processor-executable instructions;

[0055] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0056] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0057] This invention integrates video surveillance data and access control authentication data, using a time-series location point set and authentication time as input to achieve unified modeling of personnel movement trajectories and identity information. Full-process perception of personnel behavior replaces single-dimensional judgment, transforming anomaly identification from passive post-event verification to proactive real-time early warning, thus improving the adaptability and response efficiency of security systems in complex campus environments.

[0058] The identity trajectory two-layer mapping network uses trajectory units as the basic layer and identity information as the association layer. The weights of the association edges are determined based on the proximity of the authentication time and the location point time. Connectivity analysis can identify conflict structures such as trajectory breaks, identity reuse, and time inconsistencies, marking conflicting individuals as objects to be verified. The weights are determined by the actual time difference, providing a quantitative basis for conflict determination, reducing false alarms and false negatives, and providing accurate targets for subsequent control.

[0059] The access control collaborative control matrix uses access control devices as rows and time windows as columns, transforming the time distribution of the arrival of the object to be verified at each access control device into the corresponding window's restriction state. Matrix-based arrangement enables joint deployment of multiple access control devices in the time dimension, with unified generation of control strategies. Arrival times are pre-calculated based on historical trajectory units, filling in the corresponding window's restriction state to avoid isolated control and reduce missed control windows. Issued timing control commands and early warning messages are simultaneously pushed to the security terminal, collecting execution feedback from the access control devices and confirmation results from the security terminal. If execution anomalies or lack of confirmation occur, control strategies can be promptly detected and adjusted, shortening the time from anomaly detection to resolution and improving overall prevention and management efficiency. Attached Figure Description

[0060] Figure 1 This is a flowchart illustrating the campus AI security access control and early warning method according to an embodiment of the present invention;

[0061] Figure 2 A flowchart illustrating the method for constructing a two-layer identity trajectory mapping network;

[0062] Figure 3 A graph showing the combined analysis of path curvature and direction change rate for continuous trajectory segments;

[0063] Figure 4 This is a timeline trend chart for the issuance of access control collaborative control commands and conflict alarms. Detailed Implementation

[0064] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0065] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0066] Figure 1This is a flowchart illustrating the campus AI security access control and early warning method according to an embodiment of the present invention. The present invention provides a campus AI security access control and early warning method comprising:

[0067] Collect video surveillance data and access control authentication data within the campus; extract the temporal location set of personnel from the video surveillance data; and extract identity information and authentication time from the access control authentication data.

[0068] A two-layer identity trajectory mapping network is constructed. The trajectory unit is formed by clustering time-series location points as the base layer, and the association edge is established between identity information and trajectory unit as the association layer. The weight of the association edge is determined by the temporal proximity between the authentication time and the location time. Connectivity analysis is performed on the two-layer mapping network to identify conflict structures.

[0069] Mark the personnel corresponding to the conflict structure as objects to be verified, and extract the current location and historical trajectory units of the objects to be verified.

[0070] Establish an access control collaborative control matrix. The rows of the matrix correspond to access control devices, the columns correspond to time windows, and the elements represent control states. Calculate the time distribution of the object to be verified arriving at each access control device based on historical trajectory units, and fill the corresponding time windows of the matrix with restriction states.

[0071] Based on the access control collaborative control matrix, timing control commands are sent to the access control devices and early warning messages are pushed to the security terminal. The execution feedback from the access control devices and the confirmation results from the security terminal are collected.

[0072] Figure 2 This is a flowchart illustrating the method for constructing a two-layer identity trajectory mapping network. The network is constructed by clustering temporal location points into trajectory units as the base layer, and establishing association edges between identity information and trajectory units as the association layer. The weights of these association edges are determined by the temporal proximity between the authentication time and the location point time. Connectivity analysis is performed on the two-layer mapping network to identify conflict structures, including:

[0073] Incremental clustering is performed on the temporal location point set. When a location point is received, the spatiotemporal distance to the end of the existing trajectory unit is calculated. When the spatiotemporal distance meets the continuation condition, the location point is added to the existing trajectory unit. When the spatiotemporal distance does not meet the continuation condition, a new trajectory unit is created.

[0074] For the authentication time of identity information, the tolerance range is expanded on the time axis, and the trajectory unit is searched within the tolerance range. The association strength is determined according to the distance from the authentication time to the time boundary of the trajectory unit, and an association edge carrying the association strength is established between the identity information and the trajectory unit.

[0075] Perform bidirectional propagation operation on the two-layer mapping network, propagate identity labels from identity nodes to trajectory nodes, and return verification results from trajectory nodes to identity nodes. When identity labels conflict or verification results fail, set conflict markers on the associated edges.

[0076] Scan the associated edges carrying conflict markers, trace the identity nodes and trajectory nodes corresponding to the conflict markers to construct a local subnetwork, and count the out-degree of identity nodes and the in-degree of trajectory nodes in the local subnetwork.

[0077] When the in-degree of a trajectory node is multiple, it is identified as an identity competition conflict; when the out-degree of an identity node is multiple, it is identified as a trajectory splitting conflict. The conflict type is then output along with the local sub-network.

[0078] When performing incremental clustering on a time-series location point set, whenever a new location point is acquired from the video surveillance data, the spatiotemporal distance between that location point and the end nodes of all existing trajectory units is immediately calculated. The calculation of the spatiotemporal distance comprehensively considers both spatial and temporal dimensions: the spatial dimension uses Euclidean distance in the camera coordinate system, and the temporal dimension uses the difference between the location point's time and the end time of the trajectory unit. These two dimensions are then weighted and fused to obtain the comprehensive spatiotemporal distance. ,Right now ,in Indicates spatial distance. Indicates time difference, and These are spatial weighting coefficients and temporal weighting coefficients, which can be calibrated based on the actual camera deployment density and personnel movement speed in a campus setting. Less than the preset continuation threshold When the continuation condition is met, the location point is appended to the end of the corresponding trajectory unit, and the temporal boundary and spatial range of the trajectory unit are updated; when Greater than or equal to If the continuation condition is not met, a new trajectory unit is created starting from that location point. This incremental processing method avoids batch recalculation of the entire set of location points, maintaining low computational latency in scenarios with large-scale pedestrian traffic on campus, and ensuring near real-time synchronization between trajectory unit generation and video stream acquisition.

[0079] When expanding the tolerance range for the authentication time of identity information, the authentication time is used as the basis. Extend forward and backward by the center. Forming tolerance range . The value of this parameter needs to be determined by combining the response latency of the access control device and the frame rate of the camera, and is usually in the range of 2 to 5 seconds. Within this tolerance range, all trajectory units are searched, and the set of trajectory units whose time boundaries overlap with the tolerance range is selected. For each candidate trajectory unit, the authentication time is calculated. Minimum distance to the time boundary of the trajectory unit And use this to determine the strength of the association. The strength of the association is negatively correlated with distance, when... The smaller the time frame, the higher the correlation between the authentication behavior and the trajectory time, and the stronger the association. When the correlation strength is large, the correlation strength decreases. Specifically, a Gaussian attenuation method can be used. ,in The attenuation control parameter determines the attenuation rate of the association strength with distance. Based on the aforementioned association strength, a link carrying the association strength is established between the identity information node and the trajectory unit node. The directed associated edges are identified and stored in the association layer of the two-layer mapping network.

[0080] When performing bidirectional propagation on a two-layer mapping network, the propagation process consists of two phases: forward propagation and backward propagation. In the forward propagation phase, the identity node starts from the identity node and propagates the identity label along the associated edges to the connected trajectory nodes. Each associated edge carries the association strength. The confidence level is also transmitted along with the propagation. When a trajectory node simultaneously receives identity tags from multiple different identity nodes, it indicates that the trajectory unit is claimed by multiple identity information, resulting in an identity tag conflict, and a conflict marker is set on the corresponding associated edge. In the reverse propagation phase, starting from the trajectory node, the verification results of the trajectory unit's physical reachability, spatiotemporal continuity, etc., are propagated back to the identity node along the associated edge. If the verification results received by the identity node contradict its own claimed location or time, the verification result is deemed to have failed, and a conflict marker is also set on the associated edge. The bidirectional propagation operation diffuses single-point authentication anomalies into structural conflict signals at the network topology level, enabling isolated anomaly events to form identifiable conflict patterns in the network.

[0081] When scanning associated edges carrying conflict markers, the identity nodes and trajectory nodes connected to each conflicting associated edge are extracted one by one. These nodes are then used as seeds to expand outwards to include one-hop neighbors, incorporating all nodes and edges within the expansion range into the local subnetwork. The construction of the local subnetwork follows the principle of minimum closure, that is, minimizing the subnetwork size while including all conflicting nodes to avoid introducing irrelevant nodes that could interfere with subsequent statistical analysis. For each identity node in the local subnetwork, its out-degree (the number of associated edges emanating from that identity node) is calculated; for each trajectory node, its in-degree (the number of associated edges pointing to that trajectory node) is calculated. The statistical results of out-degree and in-degree directly reflect the structural characteristics of the conflict, providing a quantitative basis for subsequent conflict type determination.

[0082] When the in-degree of a trajectory node in a local subnetwork is greater than 1, it means that multiple different identities claim to be associated with that trajectory unit, i.e., multiple identities are competing for the same spatiotemporal trajectory, which is identified as an identity competition conflict. In real-world scenarios, this type of conflict corresponds to a situation where a real person's movement trajectory is claimed by multiple access control authentication records, possibly stemming from multiple people using the same access card, or from incorrect associations caused by time alignment errors between authentication records and video trajectories. When the out-degree of an identity node in a local subnetwork is greater than 1, it means that the identity information has simultaneously established associations with multiple spatiotemporally discontinuous trajectory units, i.e., a single identity appears on multiple contradictory trajectories within the same time period, which is identified as a trajectory splitting conflict. In real-world scenarios, this type of conflict corresponds to the same identity information being recorded at multiple locations where it is physically impossible for it to appear simultaneously, possibly stemming from identity misuse or abnormal duplication of access control authentication data.

[0083] After conflict type identification, the conflict type label (identity competition or trajectory splitting) and the corresponding local sub-network structure are output together as input for subsequent object verification labeling and access control collaborative control. All identity nodes involved in the local sub-network are marked as objects to be verified, and their historical trajectory units are directly provided by the trajectory nodes associated in the local sub-network, without the need to re-retrieve the full data, thus ensuring the timeliness of the entire process from conflict identification to control response. In actual campus deployment, the number of nodes and edges in the two-layer mapping network increases linearly with the scale of on-campus personnel. The combined strategy of incremental clustering and local sub-network construction ensures that the overall computational complexity is controllable, meeting the basic requirements of real-time response in campus security scenarios.

[0084] A bidirectional propagation operation is performed on the two-layer mapping network, propagating identity labels from identity nodes to trajectory nodes and returning verification results from trajectory nodes to identity nodes. When identity labels conflict or verification results fail, conflict markers are set on the associated edges, including:

[0085] Assign a propagation timestamp to each associated edge in the two-layer mapping network. The propagation timestamp identifies the propagation order of the identity label along the associated edge and assigns values ​​to the propagation timestamps according to the association strength of the associated edge.

[0086] Activate associated edges sequentially according to the order of propagation timestamps, extract identity tags from the identity nodes connected by the associated edges and propagate them to the trajectory nodes, record the binding relationship between the propagation timestamp and the identity tag when the trajectory node receives the identity tag, and store it in the trajectory node's receiving record;

[0087] For trajectory nodes with multiple binding relationships in the received records, extract the identity tags corresponding to the multiple binding relationships, calculate the attribute similarity between the identity tags, determine the tag conflict when the attribute similarity is lower than the similarity threshold, and determine the tag consistency when the attribute similarity is not lower than the similarity threshold. The determination results of tag conflict or tag consistency are then encapsulated as verification results.

[0088] The verification results are sent back to the identity node along the associated edge in reverse order of the propagation timestamp. When the identity node receives the verification results, it parses the judgment results. When the judgment results show that the labels are consistent, the verification process is completed. When the judgment results show that the labels conflict, a conflict marker is set on the corresponding associated edge.

[0089] After the two-layer mapping network is constructed, a bidirectional propagation operation needs to be performed on the network to identify conflicting structures between identities and trajectories. The bidirectional propagation consists of a forward propagation phase and a backward propagation phase. The two phases work together to detect and label potential conflicting relationships in the network.

[0090] The prerequisite for the forward propagation phase is assigning a propagation timestamp to each associated edge in the network. The purpose of the propagation timestamp is to determine the propagation order of identity tags in the network, ensuring that propagation operations on different associated edges are executed sequentially according to a predetermined priority, rather than being activated concurrently and out of order. The assignment of the propagation timestamp value is determined based on the association strength of each associated edge: the higher the association strength, the more reliable the binding relationship between the identity node and the trajectory node connected by the associated edge, thus assigning a smaller propagation timestamp value, making it activated first in the propagation sequence; the lower the association strength, the larger the propagation timestamp value, activating it later in the sequence. Let the propagation timestamp of a certain associated edge be... The correlation strength is After sorting all associated edges in descending order of association strength, then sort them according to their position. Integer numbers are assigned to create a globally ordered propagation sequence. This mechanism ensures that highly reliable identity-trajectory associations participate in propagation first, providing a more stable benchmark for subsequent conflict detection.

[0091] Each associated edge is activated sequentially in ascending order of propagation timestamps. When activating an associated edge, an identity tag is extracted from the identity node connected to that edge. The identity tag includes the person's ID number, identity category (e.g., faculty, student, visitor), and other attribute fields. After extraction, the identity tag is propagated along the associated edge to the corresponding trajectory node. Upon receiving an identity tag, the trajectory node synchronously records the currently activated propagation timestamp and establishes a binding relationship between the propagation timestamp and the identity tag, storing this in the trajectory node's reception record. The reception record is maintained as a list, with each entry containing a propagation timestamp value. The corresponding identity tag content. If a trajectory node receives only one set of binding relationships, then there is no multi-identity competition for that node in the current propagation round, and no conflict determination is required; if multiple sets of binding relationships are received, then the conflict determination process must be entered.

[0092] For trajectory nodes with multiple binding relationships in the received records, the identity tags corresponding to all binding relationships are extracted, and pairwise attribute similarity is calculated for these identity tags. The attribute similarity calculation comprehensively considers the matching degree of the identity category field, identity number field, and other structured attribute fields. Let the two identity tags be... and Their attribute similarity is denoted as For the category field, an exact match is used: if the two categories are the same, the field contributes full points; otherwise, it contributes zero points. The same exact match is used for the ID field. For other structured attribute fields, a field-level weighted average is used to calculate the similarity score. The scores of each field are then summed according to preset weights and normalized to an interval. This yields the final attribute similarity. Set the similarity threshold to... ,when When, it is determined that there is a label conflict between the two identity labels; when If two identity tags match, the system determines that the two tags are identical. If the trajectory node receives more than two identity tags, then the system calculates the identity tags for each tag pair individually. As long as there exists any pair that satisfies This means that the trajectory node is determined to have a label conflict. The determination result (label conflict or label consistency) is encapsulated into a verification result, which also records the propagation timestamps of all participants in the determination. And corresponding identity tags, so that the specific associated edge can be accurately located during subsequent feedback.

[0093] The reverse propagation phase is executed in reverse order of propagation timestamps, starting from the associated edge with the largest timestamp and sequentially propagating the verification results to associated edges with smaller timestamps. During propagation, the results are transmitted in the reverse direction of the original associated edges, from the trajectory node to the corresponding identity node. Upon receiving the verification results, the identity node parses the judgment result field: if the judgment result indicates label consistency, the association between the identity node and the trajectory node has passed verification, completing the verification process for that associated edge, and no marking is made; if the judgment result indicates label conflict, a conflict marker is set on the corresponding associated edge. The conflict marker includes the identifier of the trajectory node where the conflict occurred, the set of identity labels involved in the conflict, and the corresponding propagation timestamp set, to support subsequent conflict structure analysis and personnel marking operations.

[0094] In real-world campus security scenarios, a trajectory node corresponds to a unit of personnel movement trajectory in a certain area within a certain period of time. If the same trajectory unit is associated with multiple different identity nodes, and the personnel attributes corresponding to these identity nodes are significantly different (e.g., one is a visitor, and the other is a faculty member), then the attribute similarity... Below the threshold This triggers a tag conflict determination. Physically, this means that conflicting personnel records exist in the same area at the same time, potentially indicating inconsistencies between access control authentication data and video surveillance data, thus serving as a security early warning.

[0095] Once the conflict marking is complete, the bidirectional propagation operation ends. The associated edges carrying the conflict markings form a conflict structure in the two-layer mapping network. The identity nodes involved in the conflict structure and their corresponding personnel will be marked as objects to be verified in subsequent processes, entering the access control collaborative control and early warning processing flow. The entire bidirectional propagation mechanism, through the orderly control of propagation timestamps and the quantitative determination of attribute similarity, achieves systematic detection of potential identity conflicts in the network, providing accurate conflict location data for subsequent security responses.

[0096] The personnel corresponding to the conflict structure are marked as objects to be verified, and the current position and historical trajectory units of the objects to be verified are extracted, including:

[0097] Extract identity nodes from the conflict structure, read the identity information carried by the identity nodes to determine the corresponding personnel, assign verification priority identifiers to the personnel and mark them as objects to be verified;

[0098] Tracing back the trajectory nodes associated with the object to be verified from the conflict structure, extracting the trajectory units corresponding to the trajectory nodes to construct a trajectory set, and arranging the trajectory units in the trajectory set in time sequence according to the time boundary;

[0099] In the temporal arrangement, identify trajectory units whose time boundaries are later than the current time, extract the end position point as the current position of the object to be verified, and identify trajectory units whose time boundaries are earlier than the current time in the temporal arrangement as historical trajectory units;

[0100] For each historical trajectory unit, a set of temporal location points is extracted. Spatial clustering and temporal continuity analyses are performed on the set of temporal location points. Based on the spatial clustering analysis, the activity area type of the historical trajectory unit is labeled, and based on the temporal continuity analysis, the behavior pattern type of the historical trajectory unit is labeled.

[0101] The current location, historical trajectory unit, activity area type, and behavior pattern type of the object to be verified are associated and stored to generate a spatiotemporal behavior profile.

[0102] After identifying the conflict structure in the two-layer mapping network, it is necessary to locate and label specific personnel within the conflict structure. The conflict structure contains several identity nodes, each carrying identity information extracted from access control authentication data, including fields such as personnel number, role category, and department. Reading the identity information of an identity node uniquely identifies the corresponding campus personnel. After identifying the personnel, a verification priority identifier is assigned to each person. The priority is determined by the severity of the conflict structure: if multiple identity nodes are associated with the same trajectory node in the conflict structure, or if the weights of the associated edges differ significantly, the corresponding personnel have a higher verification priority; if only a single identity node has contradictory associations with multiple trajectory nodes in the conflict structure, the verification priority is relatively low. After priority assignment, these personnel are uniformly marked as objects to be verified. Subsequent trajectory backtracking and behavioral analysis are all carried out using objects to be verified as the processing unit.

[0103] When tracing back the trajectory nodes associated with the object to be verified from the conflict structure, the process traverses along the association edges of the association layer to the base layer, collecting all trajectory nodes that have an association edge with that identity node. Each trajectory node corresponds to a trajectory unit, which is a cluster of several temporal location points and has a clear temporal boundary, including the start time. and the end time All traced trajectory units are aggregated to construct the trajectory set of the object to be verified. Then, the trajectory set is arranged chronologically according to the time boundaries of each trajectory unit, specifically... Sort by primary key, sort by secondary key Sort the secondary sort key in ascending order so that the trajectory set presents a time sequence from early to late, which makes it easier to distinguish the historical trajectory unit from the trajectory unit to which the current position belongs.

[0104] After completing the chronological arrangement, take the current time as the starting point. Based on this, trajectory units in the time sequence are divided into two categories. All... Later The trajectory units indicate that their corresponding activities occurred after the current moment, belonging to predictive trajectory units. The endpoint of this type of trajectory unit, i.e., the position with the smallest timestamp, is extracted as the current position of the object to be verified. This processing logic is based on the fact that there is a certain delay in the acquisition and processing of video surveillance data; the time boundary of some trajectory units may be slightly later than the current system moment, but their endpoint positions are the spatial coordinates of the most recently acquired data, thus reasonably representing the actual position of the object to be verified. Earlier The trajectory units that are not identified are designated as historical trajectory units. These trajectory units record the complete activity traces of the object to be verified before the current moment and are the core data source for subsequent behavior analysis.

[0105] For each historical trajectory unit, the set of all temporal location points within that unit is extracted, and spatial clustering analysis and temporal continuity analysis are performed. The goal of spatial clustering analysis is to determine whether the location points within the trajectory unit exhibit a concentrated spatial distribution. Specifically, the spatial centroid coordinates of all location points within the trajectory unit are calculated, and then the average spatial deviation of each location point from the centroid is calculated. .like Below the preset aggregation threshold If the activity area corresponding to the trajectory unit is considered to be of the "staying-in" type, that is, the person stays continuously within a small area, and the activity area type is labeled as "staying-in area"; if Higher than If the location points are relatively scattered, the activity area type is labeled as "movement area". Based on this, we can further combine the density distribution of location points to identify whether there are multiple local clustering centers within the trajectory unit. If multiple clustering centers exist, they are labeled as "multi-point dwelling area" to distinguish between single dwelling and multi-point travel behavior.

[0106] Temporal continuity analysis focuses on the distribution characteristics of time intervals between location points within a trajectory unit. It extracts the time interval sequence between adjacent location points and calculates the mean of the time intervals. with standard deviation If the mean and standard deviation of the time intervals are small, it indicates that the location points are collected densely and evenly, corresponding to a "continuous activity" behavior pattern; if there are obvious long interval segments in the time interval sequence, that is, a certain time interval is significantly longer than... (in If the sensitivity adjustment coefficient (usually between 2 and 3) is used, it indicates that there is intermittent interruption within the trajectory unit, and the corresponding behavior pattern is labeled as "intermittent activity." If the number of location points within the entire trajectory unit is extremely small and the time span is large, it is labeled as "low-frequency occurrence," indicating that personnel only occasionally appear in the monitoring field of view during that period. Through the above two types of analysis, each historical trajectory unit obtains two labeling fields: activity area type and behavior pattern type, upgrading the trajectory unit from a simple location sequence to a behavior fragment with semantic descriptive capabilities.

[0107] After annotation, the current location of the object to be verified, all historical trajectory units, and their corresponding activity area types and behavior pattern types are associated and stored to generate a spatiotemporal behavior profile. The spatiotemporal behavior profile uses the personnel number of the object to be verified as the index key, and the profile content is organized sequentially according to the historical trajectory units arranged chronologically. Each record includes the time boundary of the trajectory unit. and The archive includes spatial coverage, activity area type annotations, behavior pattern type annotations, and the original location point set within the trajectory unit. The current location is appended as an independent field at the end of the archive, annotating the collection time and spatial coordinates. The generation of the spatiotemporal behavior archive provides structured input data for the subsequent establishment of the access control collaborative control matrix: through the historical trajectory units in the archive, the activity patterns and movement path preferences of the object to be verified can be inferred; through the activity area type and behavior pattern type, the prediction accuracy of the time distribution can be further refined, thereby making the timing control commands of the access control device more closely match the actual behavioral characteristics of the object to be verified, improving the accuracy and timeliness of security linkage response.

[0108] Establish an access control collaborative control matrix. Rows in the matrix correspond to access control devices, columns correspond to time windows, and elements represent control states. Calculate the time distribution of the object to be verified arriving at each access control device based on historical trajectory units, and fill the corresponding time windows in the matrix with restriction states, including:

[0109] Read the spatial coordinates and device number of the access control device, and assign row indexes to the access control collaborative control matrix according to the device number; set the prediction period and time window interval, divide the time window sequence and assign column indexes to generate the access control collaborative control matrix;

[0110] Extract the temporal position data of the historical trajectory units of the object to be verified, perform trajectory morphology analysis to extract path curvature and direction change rate, and identify movement habit characteristics based on path curvature and direction change rate; couple the movement habit characteristics with the current position to deduce the position evolution process of the object to be verified within the prediction period, extract evolution nodes and evolution time from the position evolution process, and combine them to generate a spatiotemporal evolution sequence;

[0111] Determine the spatial reachability from the evolution node to the access control device in the spatiotemporal evolution sequence, identify the associated access control device based on the spatial reachability, and establish a binding relationship between the device number of the associated access control device and the evolution time; locate the row index corresponding to the device number in the access control collaborative control matrix and the column index of the time window to which the evolution time belongs based on the binding relationship, fill the location with the restriction status and attach the verification priority identifier of the object to be verified.

[0112] The system reads the spatial coordinates and device numbers of access control devices deployed in various areas of the campus. Using the device number as a unique index key, it assigns row indices to the access control collaborative control matrix according to the lexicographical order of the device numbers or a preset numbering rule, ensuring a one-to-one correspondence between each row and a unique access control device. The prediction period is defined as a future time range starting from the current moment. The length of this period can be flexibly configured according to the campus security response cycle, typically covering several minutes to tens of minutes. The prediction period is evenly divided into time window sequences at fixed intervals, with each time window corresponding to a column in the matrix, and the column index increasing sequentially from 0. Each element of the access control collaborative control matrix is ​​initialized to a normal passage state, and subsequently, a restriction state is filled in at the corresponding position based on the location evolution of the object to be verified. Once the matrix structure is established, it can carry the linkage control information of multiple objects to be verified, supporting concurrent control scheduling across devices and time periods.

[0113] The temporal location data of the object to be verified is extracted from historical trajectory units. This data consists of a series of two-dimensional or three-dimensional spatial coordinates with timestamps, reflecting the object's actual movement trajectory over a period of time. Trajectory morphology analysis is performed on the temporal location data to calculate path curvature and direction change rate. Path curvature describes the degree of bending of the trajectory in space, which can be measured by the ratio of the cumulative turning angle of the broken lines formed by adjacent points to the total path length. Higher curvature indicates that the object tends to frequently turn within a region, while lower curvature indicates a relatively straight movement path. Direction change rate describes the average change in movement direction per unit time, reflecting the frequency and magnitude of directional adjustments made by the object during movement. Joint analysis of path curvature and direction change rate can identify the movement habit characteristics of the object to be verified, such as whether it tends to travel straight along fixed corridors, whether it habitually detours around specific areas, and whether it exhibits periodic back-and-forth behavior. These movement habit characteristics provide behavioral prior constraints for subsequent position evolution inference, making the prediction results closer to the actual movement patterns of the object, rather than relying solely on simple linear extrapolation.

[0114] The extracted movement habit features are coupled with the current location of the object to be verified to construct a location evolution projection process. Starting from the current location, and combining the directional preferences and speed patterns implicit in the movement habit features, the spatial location sequence that the object may traverse is progressively projected within the prediction period. During the projection process, the possible distribution area of ​​the location is calculated for each time step, and one or more candidate evolution paths are generated by comprehensively considering the statistical characteristics of historical movement speed and the laws of directional change. Key evolution nodes are extracted from the candidate paths. Evolution nodes refer to spatial locations that are representative in the location evolution process, such as turning points where the direction changes significantly, boundary points for entering new areas, or proximity points to specific functional areas. Each evolution node corresponds to an evolution time, that is, the prediction time corresponding to the location of the node. The evolution nodes are bound one by one with their corresponding evolution times to form a spatiotemporal evolution sequence. This sequence completely describes the location evolution trajectory of the object to be verified within the prediction period in the form of ordered spatiotemporal coordinate pairs.

[0115] For each evolution node in the spatiotemporal evolution sequence, determine its spatial reachability with each access control device. Spatial reachability is determined based on the Euclidean distance between the spatial coordinates of the evolution node and the spatial coordinates of the access control device, and a spatial reachability threshold is set. When the distance between the evolution node and a certain access control device does not exceed At that time, the node is deemed to have spatial reachability to the access control device. Furthermore, the reachability determination can be constrained by the physical path topology on the campus map, excluding cases where the device is close but obstructed by buildings or lacks a path, thus ensuring that the identified associated access control devices are physically reachable by the object to be verified. Access control devices that meet the spatial reachability conditions are marked as associated access control devices, and their device numbers are linked to the evolution time of the corresponding evolution node, forming a mapping pair of device numbers and evolution times for precise positioning in subsequent matrix filling operations.

[0116] Based on the device number in the binding relationship, the corresponding row index is located in the access control collaborative control matrix; based on the evolution time in the binding relationship, the time window to which that time belongs is determined, thereby locating the corresponding column index. The row index and column index together determine the unique element position in the matrix. A restriction state is filled at this position, indicating that the access control device should perform interception or forced secondary verification operations on the object to be verified within the corresponding time window, prohibiting its automatic release. Simultaneously, a verification priority identifier for the object to be verified is attached to the matrix element filled with the restriction state. The verification priority identifier is comprehensively evaluated based on the severity of the conflict structure, the degree of anomaly in the historical trajectory, and the proximity of the current location to sensitive areas. A higher priority indicates that the object requires faster intervention from security personnel. When the binding relationships of multiple objects to be verified point to the same matrix element, the highest priority identifier is retained, and the numbers of all associated objects are recorded in the supplementary information. This allows the security terminal to simultaneously obtain complete information on associated objects when receiving an alert push, supporting security personnel in making comprehensive judgments and handling decisions.

[0117] After the access control collaborative control matrix is ​​filled in, it presents a comprehensive picture of the linkage control with precise annotations in both the time and device dimensions. It reflects which access control devices need to enter the restricted state within which time windows, and also carries the priority information of the objects to be verified associated with each restricted state. This structured matrix representation allows subsequent generation of timing control commands and push of warning messages to directly use the matrix as the data source. It generates the timing control sequence for each access control device by traversing rows, and generates a warning list to be pushed to the security terminal by prioritizing it. This achieves unified driving of control logic and warning logic, effectively reducing the complexity of control coordination in multi-device linkage scenarios.

[0118] Extract the temporal position data of the historical trajectory units of the object to be verified, perform trajectory morphology analysis to extract path curvature and direction change rate, and identify movement habit features based on path curvature and direction change rate, including:

[0119] Extract temporal position data from the historical trajectory units of the object to be verified, convert the temporal position data into a spatial coordinate point sequence, and segment the spatial coordinate point sequence according to temporal continuity to form continuous trajectory segments;

[0120] A reference vector is constructed by the start and end points of a continuous trajectory segment. The vertical distance from each point on the continuous trajectory segment to the reference vector is extracted. The vertical distance is statistically processed to generate the path curvature.

[0121] Extract the movement direction of the position points on the continuous trajectory segment and convert it into an angle representation. Calculate the change in adjacent angles to construct an angle change sequence. Perform wave analysis on the angle change sequence to extract the wave amplitude as the direction change rate.

[0122] Establish a mapping table between path curvature and direction change rate. The mapping table is constructed with path curvature as the row index and direction change rate as the column index to form a two-dimensional structure. The index intersection position stores the type identifier of the movement habit feature.

[0123] Based on the path curvature and direction change rate of the object to be verified, the corresponding position is retrieved from the mapping table, and the type identifier stored at the corresponding position is read to determine the movement habit characteristics.

[0124] When extracting temporal location data from the historical trajectory units of the object to be verified, the location points recorded in the trajectory units need to be arranged by timestamps to form a spatial coordinate point sequence with temporal order. Each location point contains an x-coordinate, a y-coordinate, and a corresponding timestamp, with the coordinate system established using the reference point of the campus map as the origin. The spatial coordinate point sequence is segmented according to temporal continuity. The specific criterion is: if the time interval between two adjacent location points exceeds a preset breakpoint threshold, the segment is cut off at that point, forming an independent continuous trajectory segment. The location points within each continuous trajectory segment remain temporally continuous, representing the spatial trajectory of the object to be verified during an uninterrupted movement process. The purpose of segmentation is to avoid trajectory distortion caused by missing data or personnel stops, ensuring that subsequent morphological analysis is based on real continuous motion data.

[0125] For each continuous trajectory segment, a reference vector is constructed using the starting and ending coordinates of that segment. The reference vector represents the overall displacement direction of that trajectory segment and serves as a baseline for measuring the deviation of the actual walking path from a straight line. The perpendicular distance from each intermediate point on the continuous trajectory segment to the line containing the reference vector is extracted. This perpendicular distance reflects the degree to which the actual path deviates from the reference vector at that point. The perpendicular distances of all intermediate points are statistically processed, and the mean and maximum values ​​are calculated. The weighted combination of the mean and maximum values ​​is defined as the path curvature. ,Right now ,in The average vertical distance. This represents the maximum vertical distance. and These are the corresponding weighting coefficients, and Path curvature The larger the value, the more significant the deviation of the actual path of the trajectory relative to the line connecting the start and end points, and the more circuitous the movement path. The smaller the value, the closer the path is to a straight line, and the more direct the movement.

[0126] The movement direction between adjacent points is extracted on a continuous trajectory segment. The movement direction is calculated from the coordinate difference between adjacent points and converted into an angle representation with true north as the reference. The angle range is limited to [missing information]. Within the interval, the difference between angles in adjacent directions is calculated to obtain the angle change. To eliminate the influence of angle cycle boundaries, the change is calculated for angles exceeding... The angular change is symmetrically converted and mapped to... Within this range, an angle change sequence is constructed. Volatility analysis is performed on the angle change sequence, and the standard deviation of the angle change is calculated as the rate of change of direction. ,Right now ,in The length of the angle change sequence. For the first The change in angle, This represents the mean of the angular change sequence. The rate of change of direction. The larger the value, the more frequently the direction of movement changes within that segment of the trajectory, and the more tortuous the walking route. The smaller the value, the more stable the direction of movement and the more regular the walking route.

[0127] When establishing the mapping table between path curvature and direction change rate, the path curvature... With the rate of change of direction The data is divided into intervals. Path curvature is categorized into three levels: low curvature, medium curvature, and high curvature. The corresponding interval boundaries are determined by the quantiles of historical sample data, for example, using the 25th and 75th quantiles as the dividing points. Directional change rate is also categorized into three levels: low change, medium change, and high change, with the interval boundaries determined in the same way as for path curvature. The mapping table uses the path curvature level as the row index and the directional change rate level as the column index, constructing a 3x3 two-dimensional structure, forming 9 intersecting cells. Each cell stores a type identifier for a movement habit feature, provided by a predefined movement habit classification system.

[0128] The movement habit characteristics are categorized to cover various typical campus personnel movement patterns. A combination of low path curvature and low rate of directional change corresponds to the "straight-line rapid passage" type, indicating that the person habitually moves directly along a fixed route. A combination of low path curvature and moderate rate of directional change corresponds to the "directional patrol" type, indicating that the overall route is relatively straight but the direction is adjusted to some extent. A combination of high path curvature and high rate of directional change corresponds to the "irregular wandering" type, indicating complex movement paths, frequent changes in direction, and significant differences from normal passage patterns. Other combination types are labeled based on typical behavioral patterns observed in actual campus scenarios, ensuring a clear and unambiguous semantic correspondence between the type labels and actual movement habits.

[0129] In practical use, the path curvature is calculated based on the object to be verified. With the rate of change of direction ,Will and Each value is compared with its corresponding interval boundary to determine its level. Then, the corresponding row and column index intersection is located in the mapping table, and the type identifier stored at that location is read as the movement habit feature of the object to be verified. If the calculated value of path curvature or direction change rate falls exactly on the interval boundary, it is preferentially classified into a higher level to improve the sensitivity of identifying abnormal movement patterns.

[0130] like Figure 3 As shown in the figure, the horizontal axis represents six continuous trajectory segments (S1 to S6) divided from the historical trajectory of the object to be verified. The left vertical axis represents the path curvature C, and the right vertical axis represents the rate of change of direction R, both in degrees. The bar chart shows the path curvature value for each trajectory segment, and the line chart shows the corresponding rate of change of direction value. Among them, the curvature of segment S1 is 0.15 and the rate of change of direction is 8.2°, which is typical of fast straight-line passage; the curvature of segment S6 is as high as 0.91 and the rate of change of direction rises to 52.7°, showing obvious irregular wandering characteristics. By jointly presenting the path curvature and the rate of change of direction in the same coordinate system, the morphological differences of different trajectory segments can be intuitively compared, thereby accurately identifying the movement habit characteristics of the object to be verified, and providing behavioral prior constraints for the subsequent access control collaborative control matrix.

[0131] The identified movement habit characteristics will serve as an auxiliary basis for subsequent security judgments. Combined with the conflict structure analysis results in the identity trajectory dual-layer mapping network, they will jointly support a comprehensive risk assessment of the target. When the movement habit characteristics are "irregular wandering" or other types that significantly deviate from normal passage patterns, the corresponding warning priority will be increased, and the range of the restriction state filling in the corresponding time window in the access control collaborative control matrix will also be expanded accordingly to ensure that the passage of the target can be controlled over a larger time span. The entire trajectory morphology analysis process is based on existing temporal location data in historical trajectory units, without the need to collect additional new sensor data, demonstrating strong practicality and deployability.

[0132] Based on the access control collaborative control matrix, timing control commands are issued to the access control devices and early warning messages are pushed to the security terminal. The execution feedback from the access control devices and the confirmation results from the security terminal are collected, including:

[0133] Calculate the spatial distance between each access control device based on the current location of the object to be verified, mark the access control devices that are closer as high priority and generate a timing control instruction queue;

[0134] Control commands are sequentially retrieved from the command queue, each command is assigned a unique identifier and effective time and written into the message header, and then pushed to the access control device through the message bus;

[0135] A multi-layered early warning message containing identity information, trajectory coordinate sequence, and restricted status location distribution is constructed, encapsulated, and pushed to the security terminal. The security terminal unfolds the multi-layered message according to the hierarchy, triggering identity verification, trajectory backtracking, and policy review processes respectively. The results of each process are collected and processed independently and then merged into a confirmation result.

[0136] After receiving the instruction, the access control device establishes an execution trace based on the instruction identifier. At the effective time, it starts control and monitors the authentication event, binds the authentication event and control response into an event response pair and sends it back as execution feedback. It cross-verifies the confirmation result with the event response pair. When the processing result is inconsistent with the response status, a conflict alarm is generated and fed back to the access control collaborative control matrix for policy correction.

[0137] After the strategy is programmed according to the access control collaborative control matrix, the control intent needs to be converted into executable timing instructions and distributed to each access control device. At the same time, the warning information is pushed to the security terminal to form a complete closed-loop control link.

[0138] During the instruction generation phase, the spatial distance between the object to be verified and all access control devices on campus is calculated based on the object's current location. The spatial distance is calculated using Euclidean distance based on a planar coordinate system, taking into account both the deployment location of the access control devices and their current coordinates to obtain the spatial distance value for each access control device. ,in This is a numbered index for the access control devices. All access control devices are listed by... The devices are sorted in ascending order, and those with the shortest distances are marked as high-priority devices. The determination of high priority can be based on the campus topology; for devices located on the possible paths of the object being verified, even if the spatial distance is slightly larger, their priority weight can be appropriately increased. Based on the priority ranking, the control commands corresponding to each access control device are sequentially written into the timing control command queue. Commands at the beginning of the queue correspond to higher-priority access control devices, ensuring that control responses at key entrances and exits are triggered first.

[0139] When retrieving control instructions sequentially from the instruction queue, each instruction is assigned a globally unique identifier. ,in This is the sequence number of the instruction in the queue. A unique identifier is generated using a combination of a timestamp and a device number to ensure no identifier conflicts occur in concurrent delivery scenarios. Each instruction also needs to have its effective time written into it. The timing is determined based on the starting boundary of the corresponding time window in the access control collaborative control matrix, enabling the access control device to execute control actions at the precise time. The instruction identifier and effective time are encapsulated in the message header, while the message body contains fields such as control status (restriction or permission), target device number, and validity period. After encapsulation, the instruction is pushed to the corresponding access control device via the message bus. The message bus uses a publish-subscribe mechanism, with each access control device subscribing to a message topic bound to its own number, ensuring targeted instruction delivery and avoiding broadcast interference.

[0140] During the early warning message construction phase, a multi-layered early warning message is generated for each object to be verified. The multi-layered early warning message consists of three layers: the first layer contains the identity information of the object to be verified, including identified personnel tags, associated authentication records, and conflict structure descriptions; the second layer contains a trajectory coordinate sequence, recording the historical movement path of the object to be verified in the form of time-series location points, supporting security personnel in trajectory backtracking; the third layer contains the distribution of restricted status locations, marking the control status of each access control device within the current time window, helping security personnel intuitively understand the control coverage area. After serialization and encapsulation, the three layers of content are sent to the security terminal via a push channel. Upon receiving the multi-layered early warning message, the security terminal parses and unfolds it layer by layer, triggering the identity verification process, trajectory backtracking process, and policy review process respectively. The identity verification process retrieves personnel file databases to compare and verify the identity information in the first layer, outputting a verification conclusion; the trajectory backtracking process calls historical video clips and trajectory records to visually reproduce the second-layer coordinate sequence, outputting a backtracking report; the policy review process involves security personnel manually reviewing the third-layer restricted status distribution to confirm the rationality of the control policy and outputting review opinions. The three processes described above operate independently, each collecting and processing its own results. Ultimately, the three results are merged into a unified confirmation result, which is then sent back to the control center.

[0141] Access control equipment receives a device carrying a unique identifier After receiving the control command, establish a local system with... This is the execution trace record for the index key. During execution trace record initialization, the instruction identifier, target control status, and effective time are written. And set the execution status to "pending". When the device's internal clock reaches... When the access control device initiates a control action, it switches the access control status to the control status specified by the instruction and updates the execution status to "effective". During the control validity period, the access control device continuously monitors authentication events, recording the time of each personnel authentication attempt, the authentication result (pass or fail), and the authentication method (card swipe, face recognition, or others). Each authentication event is bound to the control response status (whether the current access control is in a restricted state) when the event is triggered, forming an event response pair. ,in For the first Authentication event log, The corresponding control response status is recorded. After all event response pairs are aggregated, they are sent back to the control center via the message bus as execution feedback data.

[0142] After receiving the execution feedback and the confirmation result from the security terminal, the control center performs cross-validation on both. The core logic of cross-validation lies in comparing the processing conclusion in the security terminal's confirmation result with the response status in the event response pair. If the security terminal's policy review conclusion indicates that a certain access control device should be in a restricted state, while the event response pair shows that the device actually performed a release action within the corresponding time window, this is considered an inconsistency, triggering a conflict alarm. The conflict alarm carries the alarm type (execution deviation or policy deviation), the access control device number involved, the corresponding time window, and the specific inconsistency content, and is fed back to the access control collaborative control matrix. Upon receiving the conflict alarm, the access control collaborative control matrix corrects the control status of the corresponding row and column elements in the matrix, regenerates the timing control instructions for the affected access control devices, and re-enters the instruction issuance process, thereby achieving dynamic adaptive correction of the policy and ensuring the continuous effectiveness of campus security management.

[0143] like Figure 4 As shown in the figure, the horizontal axis represents six time points (08:20 to 08:30) within the prediction period, divided into fixed time windows, while the vertical axis represents the number of events. Dark bars represent the number of timing control commands issued to the access control device within each time window, while light bars represent the number of execution feedbacks returned by the access control device. Both are presented side-by-side in a grouped format for easy comparison of command issuance and execution completion. Dashed lines represent the number of conflict alarms triggered by cross-validation within the same time window, with diamond-shaped markers highlighting alarm events. The period from 08:24 to 08:28 is marked with a light red semi-transparent area in the figure. During this period, commands and feedback reach their peak, and the number of conflict alarms increases simultaneously, intuitively reflecting the dynamic process of system anomalies and alarms during periods of intensive control. The overall trend fully presents the closed-loop process from command generation, execution, and feedback collection to conflict alarms in the access control collaborative control matrix, verifying the real-time linkage and adaptive policy correction capabilities of the method of this invention.

[0144] In practical deployments, the reliability of the message bus is crucial to the entire closed-loop control chain. To prevent command loss or duplicate execution, the message bus must support at least-once delivery semantics, and the access control device must respond to duplicate identifiers. When issuing instructions, duplicate delivery is ignored based on the idempotency mechanism to prevent control state from being incorrectly overwritten. Simultaneously, the effective time... The accuracy depends on the clock synchronization between the control center and the access control equipment. It is recommended to use the Network Time Protocol to periodically calibrate the clocks of all devices to keep the clock deviation within an acceptable range and ensure that the timing control commands take effect accurately at the expected time.

[0145] A second aspect of this invention provides a campus AI security access control and early warning system, comprising:

[0146] The data acquisition unit is used to collect video surveillance data and access control authentication data on campus, extract the time-series location points of personnel from the video surveillance data, and extract identity information and authentication time from the access control authentication data.

[0147] The mapping construction unit is used to construct a two-layer identity trajectory mapping network. The trajectory unit is formed by clustering time-series location points as the base layer, and the association edge is established between identity information and trajectory unit as the association layer. The weight of the association edge is determined by the temporal proximity between the authentication time and the location point time. The connectivity analysis of the two-layer mapping network is performed to identify conflict structures.

[0148] The object recognition unit is used to mark the personnel corresponding to the conflict structure as objects to be verified, and to extract the current position and historical trajectory of the objects to be verified.

[0149] The matrix establishment unit is used to establish the access control collaborative control matrix. The rows of the matrix correspond to the access control devices, the columns correspond to the time windows, and the elements represent the control status. Based on the historical trajectory unit, the time distribution of the object to be verified arriving at each access control device is calculated, and the restriction status is filled in the corresponding time window of the matrix.

[0150] The control execution unit is used to issue timing control commands to the access control equipment and push early warning messages to the security terminal according to the access control collaborative control matrix, and to collect the execution feedback of the access control equipment and the confirmation results of the security terminal.

[0151] A third aspect of the present invention provides an electronic device, comprising:

[0152] processor;

[0153] Memory used to store processor-executable instructions;

[0154] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0155] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0156] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0157] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A campus AI security access control and early warning method, characterized in that, include: Collect video surveillance data and access control authentication data within the campus; extract the temporal location set of personnel from the video surveillance data; and extract identity information and authentication time from the access control authentication data. A two-layer identity trajectory mapping network is constructed. The trajectory unit is formed by clustering time-series location points as the base layer, and the association edge is established between identity information and trajectory unit as the association layer. The weight of the association edge is determined by the temporal proximity between the authentication time and the location time. Connectivity analysis is performed on the two-layer mapping network to identify conflict structures. Mark the personnel corresponding to the conflict structure as objects to be verified, and extract the current location and historical trajectory units of the objects to be verified. Establish an access control collaborative control matrix. The rows of the matrix correspond to access control devices, the columns correspond to time windows, and the elements represent control states. Calculate the time distribution of the object to be verified arriving at each access control device based on historical trajectory units, and fill the corresponding time windows of the matrix with restriction states. Based on the access control collaborative control matrix, timing control commands are sent to the access control devices and early warning messages are pushed to the security terminal. The execution feedback from the access control devices and the confirmation results from the security terminal are collected.

2. The method according to claim 1, characterized in that, A two-layer identity trajectory mapping network is constructed. The trajectory units, formed by clustering temporal location points, serve as the base layer. The association layers are established by linking identity information with these trajectory units. The weights of the association edges are determined by the temporal proximity between the authentication time and the location point time. Connectivity analysis is performed on the two-layer mapping network to identify conflict structures, including: Incremental clustering is performed on the temporal location point set. When a location point is received, the spatiotemporal distance to the end of the existing trajectory unit is calculated. When the spatiotemporal distance meets the continuation condition, the location point is added to the existing trajectory unit. When the spatiotemporal distance does not meet the continuation condition, a new trajectory unit is created. For the authentication time of identity information, the tolerance range is expanded on the time axis, and the trajectory unit is searched within the tolerance range. The association strength is determined according to the distance from the authentication time to the time boundary of the trajectory unit, and an association edge carrying the association strength is established between the identity information and the trajectory unit. Perform bidirectional propagation operation on the two-layer mapping network, propagate identity labels from identity nodes to trajectory nodes, and return verification results from trajectory nodes to identity nodes. When identity labels conflict or verification results fail, set conflict markers on the associated edges. Scan the associated edges carrying conflict markers, trace the identity nodes and trajectory nodes corresponding to the conflict markers to construct a local subnetwork, and count the out-degree of identity nodes and the in-degree of trajectory nodes in the local subnetwork. When the in-degree of a trajectory node is multiple, it is identified as an identity competition conflict; when the out-degree of an identity node is multiple, it is identified as a trajectory splitting conflict. The conflict type is then output along with the local sub-network.

3. The method according to claim 2, characterized in that, A bidirectional propagation operation is performed on the two-layer mapping network, propagating identity labels from identity nodes to trajectory nodes and returning verification results from trajectory nodes to identity nodes. When identity labels conflict or verification results fail, conflict markers are set on the associated edges, including: Assign a propagation timestamp to each associated edge in the two-layer mapping network. The propagation timestamp identifies the propagation order of the identity label along the associated edge and assigns values ​​to the propagation timestamps according to the association strength of the associated edge. Activate associated edges sequentially according to the order of propagation timestamps, extract identity tags from the identity nodes connected by the associated edges and propagate them to the trajectory nodes, record the binding relationship between the propagation timestamp and the identity tag when the trajectory node receives the identity tag, and store it in the trajectory node's receiving record; For trajectory nodes with multiple binding relationships in the received records, extract the identity tags corresponding to the multiple binding relationships, calculate the attribute similarity between the identity tags, determine the tag conflict when the attribute similarity is lower than the similarity threshold, and determine the tag consistency when the attribute similarity is not lower than the similarity threshold. The determination results of tag conflict or tag consistency are then encapsulated as verification results. The verification results are sent back to the identity node along the associated edge in reverse order of the propagation timestamp. When the identity node receives the verification results, it parses the judgment results. When the judgment results show that the labels are consistent, the verification process is completed. When the judgment results show that the labels conflict, a conflict marker is set on the corresponding associated edge.

4. The method according to claim 1, characterized in that, The personnel corresponding to the conflict structure are marked as objects to be verified, and the current position and historical trajectory units of the objects to be verified are extracted, including: Extract identity nodes from the conflict structure, read the identity information carried by the identity nodes to determine the corresponding personnel, assign verification priority identifiers to the personnel and mark them as objects to be verified; Tracing back the trajectory nodes associated with the object to be verified from the conflict structure, extracting the trajectory units corresponding to the trajectory nodes to construct a trajectory set, and arranging the trajectory units in the trajectory set in time sequence according to the time boundary; In the temporal arrangement, identify trajectory units whose time boundaries are later than the current time, extract the end position point as the current position of the object to be verified, and identify trajectory units whose time boundaries are earlier than the current time in the temporal arrangement as historical trajectory units; For each historical trajectory unit, a set of temporal location points is extracted. Spatial clustering and temporal continuity analyses are performed on the set of temporal location points. Based on the spatial clustering analysis, the activity area type of the historical trajectory unit is labeled, and based on the temporal continuity analysis, the behavior pattern type of the historical trajectory unit is labeled. The current location, historical trajectory unit, activity area type, and behavior pattern type of the object to be verified are associated and stored to generate a spatiotemporal behavior profile.

5. The method according to claim 1, characterized in that, Establish an access control collaborative control matrix. Rows in the matrix correspond to access control devices, columns correspond to time windows, and elements represent control states. Calculate the time distribution of the object to be verified arriving at each access control device based on historical trajectory units, and fill the corresponding time windows in the matrix with restriction states, including: Read the spatial coordinates and device number of the access control device, and assign row indexes to the access control collaborative control matrix according to the device number; set the prediction period and time window interval, divide the time window sequence and assign column indexes to generate the access control collaborative control matrix; Extract the temporal position data of the historical trajectory units of the object to be verified, perform trajectory morphology analysis to extract path curvature and direction change rate, and identify movement habit characteristics based on path curvature and direction change rate; By coupling mobile habit features with the current location, the position evolution process of the object to be verified is deduced within the prediction period. Evolution nodes and evolution times are extracted from the position evolution process and combined to generate a spatiotemporal evolution sequence. Determine the spatial reachability from evolution nodes to access control devices in the spatiotemporal evolution sequence, identify associated access control devices based on spatial reachability, and establish a binding relationship between the device number of the associated access control devices and the evolution time. Based on the binding relationship, locate the row index corresponding to the device number in the access control collaborative control matrix and the column index of the time window to which the evolution time belongs, fill the restriction status at the location position and attach the verification priority identifier of the object to be verified.

6. The method according to claim 5, characterized in that, Extract the temporal position data of the historical trajectory units of the object to be verified, perform trajectory morphology analysis to extract path curvature and direction change rate, and identify movement habit features based on path curvature and direction change rate, including: Extract temporal position data from the historical trajectory units of the object to be verified, convert the temporal position data into a spatial coordinate point sequence, and segment the spatial coordinate point sequence according to temporal continuity to form continuous trajectory segments; A reference vector is constructed by the start and end points of a continuous trajectory segment. The vertical distance from each point on the continuous trajectory segment to the reference vector is extracted. The vertical distance is statistically processed to generate the path curvature. Extract the movement direction of the position points on the continuous trajectory segment and convert it into an angle representation. Calculate the change in adjacent angles to construct an angle change sequence. Perform wave analysis on the angle change sequence to extract the wave amplitude as the direction change rate. Establish a mapping table between path curvature and direction change rate. The mapping table is constructed with path curvature as the row index and direction change rate as the column index to form a two-dimensional structure. The index intersection position stores the type identifier of the movement habit feature. Based on the path curvature and direction change rate of the object to be verified, the corresponding position is retrieved from the mapping table, and the type identifier stored at the corresponding position is read to determine the movement habit characteristics.

7. The method according to claim 1, characterized in that, Based on the access control collaborative control matrix, timing control commands are issued to the access control devices and early warning messages are pushed to the security terminal. The execution feedback from the access control devices and the confirmation results from the security terminal are collected, including: Calculate the spatial distance between each access control device based on the current location of the object to be verified, mark the access control devices that are closer as high priority and generate a timing control instruction queue; Control commands are sequentially retrieved from the command queue, each command is assigned a unique identifier and effective time and written into the message header, and then pushed to the access control device through the message bus; A multi-layered early warning message containing identity information, trajectory coordinate sequence, and restricted status location distribution is constructed, encapsulated, and pushed to the security terminal. The security terminal unfolds the multi-layered message according to the hierarchy, triggering identity verification, trajectory backtracking, and policy review processes respectively. The results of each process are collected and processed independently and then merged into a confirmation result. After receiving the instruction, the access control device establishes an execution trace based on the instruction identifier. At the effective time, it starts control and monitors the authentication event, binds the authentication event and control response into an event response pair and sends it back as execution feedback. It cross-verifies the confirmation result with the event response pair. When the processing result is inconsistent with the response status, a conflict alarm is generated and fed back to the access control collaborative control matrix for policy correction.

8. A campus AI security access control and early warning system, used to implement the method as described in any one of claims 1-7, characterized in that, include: The data acquisition unit is used to collect video surveillance data and access control authentication data on campus, extract the time-series location points of personnel from the video surveillance data, and extract identity information and authentication time from the access control authentication data. The mapping construction unit is used to construct a two-layer identity trajectory mapping network. The trajectory unit is formed by clustering time-series location points as the base layer, and the association edge is established between identity information and trajectory unit as the association layer. The weight of the association edge is determined by the temporal proximity between the authentication time and the location point time. The connectivity analysis of the two-layer mapping network is performed to identify conflict structures. The object recognition unit is used to mark the personnel corresponding to the conflict structure as objects to be verified, and to extract the current position and historical trajectory of the objects to be verified. The matrix establishment unit is used to establish the access control collaborative control matrix. The rows of the matrix correspond to the access control devices, the columns correspond to the time windows, and the elements represent the control status. Based on the historical trajectory unit, the time distribution of the object to be verified arriving at each access control device is calculated, and the restriction status is filled in the corresponding time window of the matrix. The control execution unit is used to issue timing control commands to the access control equipment and push early warning messages to the security terminal according to the access control collaborative control matrix, and to collect the execution feedback of the access control equipment and the confirmation results of the security terminal.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images