CN-121999438-A - Method, equipment and medium for monitoring abnormal behaviors of posts of law enforcement officers

CN121999438ACN 121999438 ACN121999438 ACN 121999438ACN-121999438-A

Abstract

The embodiment of the application discloses a method, equipment and medium for monitoring abnormal behaviors of posts of law enforcement officers, belongs to the technical field of video data monitoring, and solves the problem of low accuracy of monitoring the posts of the law enforcement officers in the prior art. The method comprises the steps of analyzing monitoring video data corresponding to law enforcement places, identifying target law enforcement objects in video frames, extracting structural semantic elements corresponding to the target law enforcement objects in the monitoring video data, carrying out scene suitability packaging on the structural semantic elements according to a preset law enforcement place type rule base to generate supervision semantic data frames, constructing an encrypted supervision semantic stream based on the continuously generated supervision semantic data frames, transmitting the encrypted supervision semantic stream to a cloud supervision analysis platform, matching the supervision semantic data frames with a preset reference law enforcement knowledge graph in the cloud supervision analysis platform, and outputting post monitoring results corresponding to the target law enforcement objects according to matching results.

Inventors

LIU FANG
HU JIABAO
ZHANG GUANGMING
WANG HAOQING
ZHANG PANPAN

Assignees

山东东沃信息技术有限公司

Dates

Publication Date: 20260508
Application Date: 20260130

Claims (10)

1. A method for monitoring abnormal behavior of a law enforcement officer, the method comprising: analyzing the monitoring video data corresponding to the law enforcement place, and identifying a target law enforcement object in the video frame, wherein the target law enforcement object at least comprises one of a target law enforcement personnel and a target law enforcement facility; Extracting a structural semantic element corresponding to the target law enforcement object from the monitoring video data, wherein the structural semantic element at least comprises one of an identity of a law enforcement officer, action classification of the law enforcement officer, a state and a position of a managed facility and a spatial topological relation among the target law enforcement objects; Performing scene suitability packaging on the structural semantic elements according to a preset law enforcement place type rule base to generate supervision semantic data frames; Based on the continuously generated supervision semantic data frames, an encrypted supervision semantic stream is constructed and transmitted to a cloud supervision analysis platform, so that the cloud supervision analysis platform matches the supervision semantic data frames with a preset reference law enforcement knowledge graph, and a post monitoring result corresponding to the target law enforcement object is output according to the matching result.
2. The method for monitoring abnormal behavior of law enforcement officers according to claim 1, wherein the analyzing the monitoring video data corresponding to the law enforcement places, and identifying the target law enforcement officers in the video frame, specifically comprises: Performing visual feature extraction on video frames in the monitoring video data, and performing similarity calculation on the extracted visual features and preset law enforcement object text description features to screen out a target law enforcement object candidate set and confidence level thereof based on calculation results; taking the candidate law enforcement objects in the target law enforcement object candidate set as nodes, and taking the position relationship and interaction relationship between the candidate law enforcement objects as edges to construct a law enforcement scenario relationship graph; Matching a corresponding relation constraint rule set in a preset law enforcement scenario common knowledge rule base based on the object types of all nodes in the law enforcement scenario relation graph; Verifying the law enforcement scenario relationship graph based on the relationship constraint rule set; And correcting the confidence coefficient of the candidate law enforcement object according to the verification result, and taking the candidate law enforcement object with the corrected confidence coefficient meeting a preset threshold as the target law enforcement object.
3. The method for monitoring abnormal behavior of law enforcement officials according to claim 1, wherein the extracting, in the monitoring video data, the structured semantic elements corresponding to the target law enforcement officials specifically comprises: aiming at the object type and the current law enforcement scene information corresponding to each target law enforcement object respectively, determining corresponding reference visual features in a preset law enforcement feature rule base; Replacing the monitoring features corresponding to the current target law enforcement object with the reference visual features, and keeping the monitoring features of other target law enforcement objects in the scene unchanged; Determining the difference degree between the corresponding inverse fact feature representation after feature replacement and the actually monitored visual feature representation, and obtaining the abnormal deviation degree corresponding to the target law enforcement object; And sorting the target law enforcement objects based on the abnormal deviation, screening candidate law enforcement objects based on sorting results, and extracting structural semantic elements of the candidate law enforcement objects.
4. The method for monitoring abnormal behavior of law enforcement officers according to claim 1, wherein the scene suitability packaging is performed on the structural semantic elements according to a preset law enforcement place type rule base to generate supervision semantic data frames, and the method specifically comprises the following steps: taking the structural semantic elements as input, and carrying out multi-hop related sub-graph retrieval in a preset law enforcement knowledge graph to generate a target context sub-graph corresponding to the semantics of the current law enforcement scene; Encoding the structural semantic elements into current scene state vectors, and performing similarity comparison with the scene state vectors cached at the previous moment; and when the similarity is smaller than a preset event trigger threshold, packaging the current scene state vector, the difference element causing the state change and the target context sub-spectrum to generate the supervision semantic data frame.
5. The method for monitoring abnormal behavior of law enforcement officers according to claim 4, wherein the encapsulating the current scene state vector, the difference element causing the state change, and the target context sub-spectrum to generate the supervision semantic data frame specifically comprises: Determining supervision value scores corresponding to the structural semantic elements respectively based on historical law enforcement data and current law enforcement scene information; Dividing the structural semantic elements into a plurality of level sets based on the supervision value scores, and differentially encoding the structural semantic elements in different level sets; Constructing a layered data packet structure based on the encoded data, wherein the layered data packet structure at least comprises a core layer containing element foundation representation and an enhancement layer containing element corresponding data; In the packaging transmission process, monitoring the supervision value score of each structural semantic element, and triggering a historical data backtracking instruction of the change element under the condition that the score change value does not accord with a preset score change threshold value so as to generate a supplement data packet based on the historical data for packaging and supplementing transmission.
6. The law enforcement officer post abnormal behavior monitoring method of claim 1, wherein the constructing an encrypted supervision semantic stream based on the continuously generated supervision semantic data frames specifically comprises: Encrypting the supervision semantic data frame by presetting a first encryption key to generate a full ciphertext; Encrypting key feature fields appearing in the supervision semantic data frame through a preset second encryption key to generate a search ciphertext and generating a semantic fingerprint based on the key feature fields, wherein the key feature fields at least comprise one of a behavior intention label field, a spatial relationship exception representation field and a facility state exception encoding field; Dividing the full ciphertext into a plurality of data slices, and generating corresponding dynamic watermarks for each data slice; based on the current moment and the equipment coding for encryption, generating a random confusion factor, and linearly transforming the ciphertext slice with the watermark through the random confusion factor to generate a coded ciphertext slice; And packaging the encoded ciphertext slice, the search ciphertext and the semantic fingerprint to generate a transmission unit, and constructing the encrypted supervision semantic stream based on the continuous transmission unit.
7. The method for monitoring abnormal behavior of law enforcement officers according to claim 6, wherein after the constructing an encrypted supervision semantic stream based on the continuously generated supervision semantic data frames and transmitting the encrypted supervision semantic stream to a cloud supervision analysis platform, the method further comprises: The semantic fingerprints received are matched with a preset unqualified law enforcement behavior feature library in a cloud supervision and analysis platform; under the condition that the matching is successful, performing multiparty calculation based on the first secret share corresponding to the cloud supervision and analysis platform and the node with the second secret share, and acquiring a result token; If the indication content corresponding to the result token is that no deep processing is needed, the encoded ciphertext slices are imported into an encryption storage system for archiving; And if the indication content corresponding to the token is the indication content which needs to be deeply processed, acquiring decryption authorization from an authorized party based on the result token, and decrypting the encoded ciphertext slice by using the decryption authorization to obtain a complete supervision semantic data frame.
8. The method for monitoring post abnormal behaviors of law enforcement officers according to claim 1, wherein the cloud supervision analysis platform matches the supervision semantic data frame with a preset reference law enforcement knowledge graph and outputs post monitoring results corresponding to the target law enforcement officers according to the matching results, specifically comprising: Matching the supervision semantic data frame with rule nodes in the preset reference law enforcement knowledge graph, and obtaining law enforcement post behavior scores based on a matching result; determining a judging threshold corresponding to the abnormal law enforcement behaviors based on the law enforcement scenario information and the historical behavior record of the target law enforcement object; comparing the law enforcement action score with the judgment threshold, and outputting law enforcement action abnormal action information if the law enforcement action score is larger than the judgment threshold; retrieving supportive and wealth pattern information associated with a current law enforcement scenario in the preset reference law enforcement knowledge pattern; and outputting the post monitoring result based on the law enforcement post abnormal behavior information and the retrieved map information.
9. A law enforcement personnel post anomaly behavior monitoring device, comprising a memory for storing computer program instructions and a processor for executing the program instructions, wherein the computer program instructions, when executed by the processor, trigger the device to perform the method of any one of claims 1-8.
10. A non-transitory computer storage medium storing computer executable instructions, wherein the computer executable instructions are capable of performing the method of any one of claims 1-8.

Description

Method, equipment and medium for monitoring abnormal behaviors of posts of law enforcement officers Technical Field The application relates to the technical field of video data monitoring, in particular to a method, equipment and medium for monitoring abnormal behaviors of posts of law enforcement personnel. Background Objective and efficient supervision of law enforcement personnel post performance is an important link for guaranteeing law enforcement activity standardization. In the prior art, video is recorded and stored through a camera arranged at a law enforcement place, and a supervisor needs to regularly or irregularly review historical videos for manual checking, or manually identify possible improper track job conditions through a real-time video round robin mode. However, the manual inspection and video review mode is limited by human resources and the attention capacity of monitoring staff, and because of the limited number of monitoring staff, the fine screening of all video contents is difficult, and only a sampling inspection mode can be adopted, so that a large number of video contents are in a state of not being effectively monitored, and a monitoring blind area is formed. Secondly, because the manual work is easily influenced by fatigue factors, the continuous attention to the video picture is easy to generate visual fatigue and attention decay, and the condition of missed detection on abnormal behaviors can occur, so that the accuracy of monitoring the posts of law enforcement personnel is lower. Disclosure of Invention The embodiment of the application provides a method, equipment and medium for monitoring abnormal behaviors of a law enforcement officer post, which are used for solving the technical problems that the prior art is limited by human resources and the attention capacity of monitoring officers, so that the accuracy of monitoring the law enforcement officer post is lower. The embodiment of the application adopts the following technical scheme: The embodiment of the application provides a law enforcement personnel post abnormal behavior monitoring method. The method comprises the steps of analyzing monitoring video data corresponding to a law enforcement place, identifying target law enforcement objects in video frames, wherein the target law enforcement objects at least comprise one of target law enforcement personnel and target law enforcement facilities, extracting structural semantic elements corresponding to the target law enforcement objects in the monitoring video data, wherein the structural semantic elements at least comprise one of identity marks of law enforcement personnel, action classifications of law enforcement personnel, states and positions of managed facilities and space topological relations among the target law enforcement objects, performing scene suitability packaging on the structural semantic elements according to a preset law enforcement place type rule base to generate supervision semantic data frames, constructing encrypted supervision semantic streams based on continuously generated supervision semantic data frames, transmitting the encrypted supervision semantic streams to a cloud supervision analysis platform, matching the supervision semantic data frames with preset reference law enforcement knowledge maps in the cloud supervision analysis platform, and outputting post monitoring results corresponding to the target law enforcement objects according to matching results. The method comprises the steps of analyzing monitoring video data corresponding to a law enforcement place, identifying target law enforcement objects in video frames, extracting visual features of the video frames in the monitoring video data, carrying out similarity calculation on the extracted visual features and preset law enforcement object text description features to screen out target law enforcement object candidate sets and confidence coefficients thereof based on calculation results, constructing a law enforcement scene relation graph by taking candidate law enforcement objects in the target law enforcement object candidate sets as nodes and taking the position relations and interaction relations between the candidate law enforcement objects as edges, matching corresponding relation constraint rule sets in a preset law enforcement scene rule base based on object types of all nodes in the law enforcement scene relation graph, checking the law enforcement scene relation graph based on the relation constraint rule sets, correcting the confidence coefficients of the candidate law enforcement objects according to check results, and taking the candidate law enforcement objects with corrected confidence coefficients meeting preset thresholds as target law enforcement objects. In one implementation mode of the application, in monitoring video data, the method extracts the structural semantic elements corresponding to the target law enforcement objects, specifically comprises the