CN-122024149-A - Vehicle sealing state detection tracing system and method based on fusion target detection

CN122024149ACN 122024149 ACN122024149 ACN 122024149ACN-122024149-A

Abstract

The invention provides a vehicle sealing state detection and tracing system based on fusion target detection, which belongs to the technical field of computer data processing and is used for actively identifying and deeply tracing abnormal events. According to the invention, potential logic contradiction is found through cross verification among multiple information, when deviation occurs in part of identification links, final error judgment can be effectively avoided, decision accuracy and operation robustness are improved, a traceable evidence chain containing complete reasoning basis is packaged, deep transparency of a decision process is realized, finally stored data is not only a simple result record, but also can be configured in a small program, a user can complete shooting, uploading and identification all-flow operation through a mobile phone without installing a special APP, various user groups are adapted, automatic identification replaces manual verification, and the sealing detection time of a single vehicle is shortened to 1 minute from 3-5 minutes in the prior art, so that the operation efficiency is greatly improved.

Inventors

LIANG CHAO
ZHANG DAOXIANG
DU SHIYONG
LIAO YONGHONG
DENG NINGFA
LI YAHE

Assignees

佛山市明睿达科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260120

Claims (10)

1. A vehicle sealing state detection traceability system based on converged target detection, the system comprising a server, a front end applet in communication with the server, and a detection terminal disposed at the server, the system being configured to include: the multi-mode feature extraction module is configured to extract visual features in parallel based on the sealing part image, and generate a visual feature set, wherein the visual feature set comprises target perception features and text recognition results; The dynamic knowledge graph construction module is configured to receive the visual feature set and the operation request data and execute the operation of constructing the multidimensional association graph; the cross-validation reasoning engine is configured to call a reasoning rule set for contradiction analysis to validate the multidimensional association graph and generate a hierarchical decision instruction; And the traceable evidence chain packaging module is configured to receive the hierarchical decision instruction, the visual feature set and the operation request data, and execute the operation of forming and storing the traceable evidence chain.
2. The fusion object detection-based vehicle encapsulation state detection trace back system of claim 1, wherein the dynamic knowledge-graph construction module is configured to perform the steps of receiving the visual feature set and the operation request data, and performing the steps of constructing a multidimensional association graph: responding to an image acquisition instruction of a client, and acquiring an image of a sealing part shot for a target sealing part; Acquiring expected seal identification information associated with a current transportation order from a service system; Acquiring scene information of the operation, wherein the scene information comprises an operation time stamp, an operation geographic position and an operator identifier; and packaging the expected seal identification information and the scene information into the operation request data and associating the operation request data with the sealing part image.
3. The vehicle sealing state detection traceability system based on fusion target detection according to claim 2, wherein the dynamic knowledge-graph construction module extracts visual features in parallel based on the sealing part image, and generating the visual feature set includes: processing the sealing part image through a target detection model, identifying at least a suspected seal area, and outputting coordinate information of the suspected seal area and a corresponding target visual feature vector; performing full-image character recognition on the sealing part image through a text recognition model, extracting all text blocks and contents thereof contained in the image, and generating a text recognition result; And integrating the target visual feature vector, the coordinate information and the text recognition result to generate the visual feature set.
4. The vehicle sealing state detection traceability system based on fusion target detection according to claim 3, wherein the dynamic knowledge-graph construction module is configured to construct a multidimensional association graph representing the relationship of the internal elements of the present operation according to the visual feature set and the operation request data: Taking the suspected seal area, the text block and the expected seal identification information and the scene information in the operation request data in the visual feature set as nodes; determining a spatial proximity relation between the suspected seal area and the text block based on the coordinate information, and generating a first association edge; Establishing a matching relation between the suspected seal area and the expected seal identification information, and generating a second association edge; Establishing a space-time generic relation between the suspected seal area and the scene information, and generating a third association edge; And the nodes, the first association edge, the second association edge and the third association edge form the multidimensional association map.
5. The vehicle sealing state detection trace back system based on fusion target detection of claim 1, wherein the cross-validation inference engine is configured to invoke a set of inference rules for contradictory analysis to validate the multi-dimensional correlation map, generating hierarchical decision instructions comprises: Traversing the multidimensional association graph, and checking whether visual internal contradiction, visual and business contradiction or time sequence and business contradiction exist according to the reasoning rule set; if no contradiction is found, generating a first type of decision instruction which represents that the sealing verification passes; If only the visual contradiction and the business contradiction are found, generating a second class of decision instructions for verifying the characterization number; If at least one of the visual contradiction or the time sequence and business contradiction is found, a third type of decision instruction representing the high risk early warning is generated, wherein the third type of decision instruction comprises a trigger command requiring to start additional verification.
6. The vehicle encapsulation state detection trace back system based on fusion target detection of claim 5, wherein the trace-source evidence chain encapsulation module is configured to receive the hierarchical decision instruction, the visual feature set, and the operation request data, and perform the operations of forming and storing a traceable evidence chain comprising: Packaging the hierarchical decision instruction, the snapshot of the multidimensional association graph according to which the instruction is generated, the visual feature set and the operation request data into a structured evidence data packet; storing the evidence data packet into a storage medium with tamper-proof characteristics to form the traceable evidence chain; and feeding back the core conclusion of the hierarchical decision instruction to the client.
7. The vehicle sealing state detection trace back system based on fusion target detection according to claim 5, wherein the trace-source evidence chain packaging module is configured to include, after generating the third type of decision instruction: Sending an additional verification request to the client to obtain a dynamic image or a multi-angle image containing the target sealing part; Updating the visual feature set based on the dynamic image or the multi-angle image to generate an enhanced visual feature set; And re-executing the steps of constructing the multidimensional association graph and subsequent contradiction analysis and verification by utilizing the enhanced visual feature set.
8. The vehicle sealing state detection trace back system based on fusion target detection according to claim 4, wherein the inference rule set includes a first rule for detecting a visual contradiction, the first rule specifically being: obtaining a standard seal characteristic template obtained by extracting characteristics of a combined seal image; Calculating the similarity between the target visual feature vector of the suspected seal area node and the standard seal feature template; And if the similarity is lower than a first threshold value and text block nodes exist in the spatial adjacent relation of the suspected seal area nodes, judging that the visual contradiction exists.
9. The vehicle sealing state detection trace back system based on fusion target detection according to claim 7, wherein the front end applet further comprises a deployment and initialization of a client module: Deploying a sealing identification client program in a front end applet, embedding a user registration login control, an image acquisition control, an HTTP request encapsulation control and a result display control, and setting calling triggering conditions of the controls; When a user triggers a client program starting instruction, the client loads basic configuration comprising a server interface address and image compression parameters, and initializes each function control to a state to be called; the client calls a user authorization interface, and when a user triggers a login control, a login page is displayed; The client generates a user identity Token instruction, and after the user completes login verification, the user identity Token instruction is stored in a local cache for identity authentication of subsequent requests.
10. The automatic detection and tracing method for the sealing state of the vehicle based on the fusion target detection is applied to the detection and tracing system for the sealing state of the vehicle based on the fusion target detection as claimed in any one of claims 1 to 9, and is characterized by comprising the following steps: acquiring an image of a sealing part sent by a client and operation request data bound with the image; extracting visual features in parallel based on the sealing part image, and generating a visual feature set, wherein the visual feature set comprises target perception features and text recognition results; Constructing a multidimensional association map representing the relationship of the internal elements of the operation according to the visual feature set and the operation request data; invoking an inference rule set for contradiction analysis to verify the multidimensional association graph to generate a hierarchical decision instruction; And forming and storing a traceable evidence chain according to the hierarchical decision instruction, the visual characteristic set and the operation request data.

Description

Vehicle sealing state detection tracing system and method based on fusion target detection Technical Field The invention relates to the technical field of computer data processing, in particular to a vehicle sealing state detection traceability system based on fusion target detection. Background Vehicle sealing is a key measure for guaranteeing safety and compliance of cargoes in links such as transportation, storage, customs clearance and the like. The accurate detection of the sealing state and the accurate recording of the sealing information are critical to the integrity of a logistics chain, and along with the development of digital technology, the field gradually evolves from traditional manual visual inspection and paper recording to the automation and digitization direction based on the image recognition technology. The digitalized vehicle sealing state detection scheme generally adopts a linear image processing flow, a target detection algorithm is utilized to locate the region of a suspected seal in the acquired image, then the region is cut out, an optical character recognition algorithm is called to perform character recognition on the cut-out small image so as to acquire the seal number, and the system stores the recognized number and information such as time, place and the like into a database to complete one-time digitalized record. In practical application, a linear data processing flow has inherent vulnerability, errors of any link of front-end target detection or subsequent character recognition are transmitted without verification, so that an error visual recognition process of a final recorded result is directly independent of data such as dispatch, circulation and the like in a business system, a mechanism for carrying out multi-source information fusion and verification in the processing process is lacked, an automatic decision process is difficult to effectively trace back and copy when disputes occur, and an improvement space exists. Disclosure of Invention The embodiment of the invention provides a vehicle sealing state detection and tracing system based on fusion target detection, which is used for extracting multi-mode features in parallel, dynamically constructing a correlation map and carrying out a digital data processing mode of cross verification reasoning, and can systematically improve the accuracy and the robustness of decision making and realize the active identification and the deep tracing of abnormal events. In order to achieve the above purpose, the invention adopts the following technical scheme: In a first aspect, a vehicle sealing state detection traceability system based on fusion target detection is provided, the system including a server, a front end applet in communication with the server, and a detection terminal disposed at the server, the system being configured to further include: the multi-mode feature extraction module is configured to extract visual features in parallel based on the sealing part image, and generate a visual feature set, wherein the visual feature set comprises target perception features and text recognition results; The dynamic knowledge graph construction module is configured to receive the visual feature set and the operation request data and execute the operation of constructing the multidimensional association graph; the cross-validation reasoning engine is configured to call a reasoning rule set for contradiction analysis to validate the multidimensional association graph and generate a hierarchical decision instruction; And the traceable evidence chain packaging module is configured to receive the hierarchical decision instruction, the visual feature set and the operation request data, and execute the operation of forming and storing the traceable evidence chain. Optionally, the dynamic knowledge graph construction module is configured to perform a method for receiving the set of visual features and the operation request data, and to perform the constructing a multidimensional association graph: responding to an image acquisition instruction of a client, and acquiring an image of a sealing part shot for a target sealing part; Acquiring expected seal identification information associated with a current transportation order from a service system; Acquiring scene information of the operation, wherein the scene information comprises an operation time stamp, an operation geographic position and an operator identifier; and packaging the expected seal identification information and the scene information into the operation request data and associating the operation request data with the sealing part image. Optionally, the dynamic knowledge graph construction module extracts visual features in parallel based on the sealing part image, and generating the visual feature set includes: processing the sealing part image through a target detection model, identifying at least a suspected seal area, and outputting coordinate information of the susp