A method and system for automatic auditing and authenticity checking of logistics invoice information
By employing a collaborative architecture of lightweight OCR small models and multimodal large models, the problems of low accuracy in identifying logistics weighbridge information and insufficient verification of document authenticity have been solved, enabling efficient and reliable automatic review and authenticity verification of logistics weighbridges.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- INNER MONGOLIA TRANSPORTATION GROUP DIGITAL LOGISTICS TECHNOLOGY CO LTD
- Filing Date
- 2026-03-10
- Publication Date
- 2026-06-05
AI Technical Summary
Existing technologies for identifying logistics weighbridge information suffer from low accuracy, difficulty in handling unstructured weighbridges with inconsistent styles and field naming, and a lack of proactive verification mechanisms for document authenticity, resulting in low efficiency, high error rates, and weak risk control capabilities.
Lightweight OCR is used to recognize small models for preprocessing, and multimodal understanding is combined with large models for correction and completion. Logical compliance and authenticity verification are performed by combining a preset audit rule base and external authentication interfaces to generate the final audit conclusion.
It significantly improved the accuracy of identifying key fields on weighbridge slips, enhanced the reliability of audit conclusions and the automation level of the system, reduced the need for manual intervention, and strengthened risk control capabilities.
Smart Images
Figure CN122157277A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of logistics and transportation, and in particular to a method and system for automatic verification and authenticity checking of logistics weighbridge information. Background Technology
[0002] In modern logistics and transportation, weighing goods is a core step in settlement, billing, and supervision. Weigh slips, as crucial evidence of weighing data, are typically in the form of paper printouts, handwritten supplements, or electronic screenshots, and are widely used in industries such as mining, ports, building materials, and energy. However, current weigh slip management still heavily relies on manual review. Staff must meticulously check key fields such as license plate number, gross weight, tare weight, net weight, and time, and determine if there are any issues such as tampering, forgery, or logical inconsistencies. This model suffers from significant problems including low efficiency, susceptibility to errors, high costs, and difficulty in preventing fraudulent activities.
[0003] In recent years, some companies have attempted to introduce OCR (Optical Character Recognition) technology to automate the extraction of weighbridge information. However, in practical applications, it has been found that traditional OCR solutions face two major challenges: First, there are many different styles of weighbridge slips, and there is a lack of a unified standard.
[0004] The weighbridge slip templates used by different regions, companies, and even weighbridge stations vary greatly—some use a three-part format, while others have free layouts; some fields are left-aligned, while others are nested within tables; some contain a mix of elements such as QR codes, barcodes, stamps, and signatures. This unstructured nature makes it difficult for general OCR models to accurately locate key fields, easily leading to omissions, misalignments, or mismatches.
[0005] Second, the naming of key fields is inconsistent, making semantic understanding difficult.
[0006] The same type of information may be expressed using multiple names on different weighbridge slips. For example, "gross weight" may be labeled as "total weight," "gross weight," or "weight before weighing"; "tare weight" may also be written as "empty vehicle weight" or "tare." Furthermore, some fields use abbreviations, dialects, or industry jargon, further increasing the difficulty of machine recognition and structured mapping. Methods based solely on keyword matching are highly susceptible to recognition failure due to naming variations.
[0007] At the same time, existing systems generally lack proactive verification mechanisms for the authenticity of documents, can only complete formal reviews, and cannot identify the use of forged ID cards, driver's licenses or operating permits, resulting in weak risk control capabilities.
[0008] Although multimodal large models (such as Qwen-VL and DeepSeek-VL) have demonstrated strong contextual reasoning capabilities in image and text understanding in recent years, their training data mainly comes from general image and text pairs on the Internet. As a result, they perform poorly in the vertical field of logistics weighbridge slips, especially in handwritten text recognition tasks. They often have problems such as misreading numbers (e.g., recognizing "6" as "0") and Chinese character recognition errors. When used directly for end-to-end recognition, the accuracy rate is less than 70%, which cannot meet the actual business needs. Summary of the Invention
[0009] The purpose of this application is to provide a method and system for automatic verification and authenticity checking of logistics weighbridge information, so as to solve the problem of low accuracy in identifying logistics weighbridge information.
[0010] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a method for automatically verifying and confirming the authenticity of logistics weighbridge information, including: Receive the weighbridge image and preprocess it to obtain a preprocessed image; The preprocessed image is input into a lightweight OCR recognition model for text recognition, and the output is a preliminary recognition result containing text content, location coordinates and confidence level. The preprocessed image and the preliminary recognition result are jointly input into the multimodal understanding model. Based on visual features, text semantics and logical relationships between fields, the preliminary recognition result is corrected, completed and semantically normalized, and structured bill of lading data is output. The structured weighbridge data is logically validated based on a pre-defined audit rule base to determine the logical compliance validation result. The external official authentication interface is called to verify the authenticity of the transportation document information contained in the structured weighbridge data, and the verification result is determined. The final audit conclusion is generated by combining the logical compliance verification results with the authenticity verification results.
[0011] Secondly, this application provides an automatic verification and authenticity check system for logistics weighbridge information, including: The image preprocessing module is used to receive the weighbridge image and preprocess it to obtain a preprocessed image; A dual-layer recognition engine, including: A lightweight OCR recognition submodule is used to perform text recognition on preprocessed images and output preliminary recognition results containing text content, location coordinates, and confidence level. The multimodal understanding submodule is used to receive the preprocessed image and the preliminary recognition result, and to correct, complete and semantically normalize the preliminary recognition result based on visual features, text semantics and logical relationships between fields, and output structured bill of lading data. The rules engine module is used to load a preset audit rules library and perform logical compliance verification on the structured weighbridge data to determine the logical compliance verification result. The external interface proxy module is used to call external official certification interfaces to verify the authenticity of transportation documents and determine the verification result. The review decision module is used to generate a final review conclusion by combining the logical compliance verification results and the authenticity verification results.
[0012] According to the specific embodiments provided in this application, this application has the following technical effects: First, the lightweight OCR recognition model provided in this application possesses high-precision recognition capabilities for key information such as handwritten Arabic numerals and Chinese characters. Inputting the pre-processed image into the lightweight OCR recognition model for text recognition effectively overcomes problems such as digit misreading (e.g., misidentifying "6" as "0") and Chinese character recognition errors caused by insufficient training of the general multimodal understanding model in logistics vertical scenarios, providing a high-confidence preliminary recognition result for subsequent processing. Second, the pre-processed image and the preliminary recognition result (including text content, location coordinates, and confidence level) are jointly input into the multimodal understanding model. This allows the multimodal understanding model to obtain additional structured weighbridge data based on visual feature analysis, guiding it to focus on key areas for contextual reasoning. This accurately corrects recognition errors of the lightweight OCR recognition model (e.g., character misjudgment), completes missing fields, and performs semantic consistency verification based on the logical relationships between fields, avoiding logical contradictions caused by isolated recognition. Finally, based on the preset audit rule base and by calling the external official certification interface, logical compliance verification and authenticity verification are performed to generate the final audit conclusion. This application significantly improves the recognition accuracy of key fields of the weighbridge slip by constructing a two-layer architecture of "lightweight OCR recognition small model preprocessing + multimodal understanding large model collaborative fine recognition", thereby improving the recognition accuracy of logistics weighbridge slip information and making the final audit conclusion more reliable. Attached Figure Description
[0013] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0014] Figure 1 A flowchart illustrating an automatic verification and authenticity check method for logistics weighbridge information provided in one embodiment of this application; Figure 2 This is a schematic diagram of the computer system execution flow provided in an embodiment of this application. Detailed Implementation
[0015] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0016] To make the objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0017] like Figure 1 As shown in the embodiment of this application, an automatic verification and authenticity check method for logistics weighbridge information is provided, including: S1: Receive the weighbridge image and preprocess it to obtain a preprocessed image.
[0018] S2: Input the preprocessed image into a lightweight OCR recognition model for text recognition, and output preliminary recognition results including text content, location coordinates and confidence level.
[0019] S3: Input the preprocessed image and the preliminary recognition result into the multimodal understanding model, and correct, complete and semantically normalize the preliminary recognition result based on visual features, text semantics and logical relationships between fields, and output structured bill of lading data.
[0020] S4: Perform logical compliance verification on the structured weighbridge data based on the preset audit rule base, and determine the logical compliance verification result.
[0021] S5: Call the external official authentication interface to verify the authenticity of the transportation document information contained in the structured weighbridge data, and determine the verification result.
[0022] S6: Combine the logical compliance verification results with the authenticity verification results to generate the final audit conclusion.
[0023] In an exemplary embodiment, the execution entity of S1 is: the image preprocessing module (running on a cloud server).
[0024] The system retrieves weighbridge slip images uploaded by users (supporting JPG / PNG formats). These images undergo standardized preprocessing, including image denoising, contrast enhancement, and automatic rotation correction.
[0025] S1 can improve the input quality of subsequent recognition models and reduce noise interference, and is especially suitable for situations such as tilted shooting angles and uneven lighting.
[0026] In an exemplary embodiment, the execution entity of S2 is a lightweight OCR recognition engine (running on a cloud server).
[0027] The preprocessed weighbridge slip image is input into a lightweight OCR recognition model for recognition. This lightweight OCR recognition model is based on the open-source PP-OCRv5 model of PaddleOCR 3.3.0 and is specifically trained and fine-tuned for logistics weighbridge slip scenarios.
[0028] The PP-OCRv5 model is trained using synthetic and real data, and consists of two training phases: detection training and recognition training. First, the detection model is trained, and then the recognition model is trained using the text regions output by the detection model.
[0029] This lightweight OCR recognition model architecture for small models consists of two core parts: Text detection module: CSPDarknet backbone network + DBNet (Differentiable Binarization) algorithm.
[0030] Text recognition module: MobileNet backbone network + SVTR (Single-Branch Visual Text Recognition).
[0031] Output: Initially identified text content; location coordinates and confidence score (0~1) for each field; The role of S2: As a pre-processor for large models, it provides high-quality intermediate recognition results and makes up for the shortcomings of large models in vertical fields, especially in handwriting recognition.
[0032] In one exemplary embodiment, the lightweight OCR recognition model is fine-tuned using a dedicated training dataset; the dedicated training dataset contains handwritten weighing slip samples from logistics scenarios, which enhances the recognition capability of the lightweight OCR recognition model for handwritten Arabic numerals and Chinese characters.
[0033] In an exemplary embodiment, S3 specifically includes: S31: The preliminary recognition results are superimposed onto the preprocessed image in the form of annotations to form an annotated composite input image.
[0034] S32: Input the composite input image into the multimodal understanding model to guide the multimodal understanding model to perform contextual reasoning and result optimization based on the annotation.
[0035] In practical applications, S3 combines the output of the small model with the preprocessed image and inputs it into a multimodal large model for collaborative and precise recognition.
[0036] The executor of S3 is a multimodal understanding engine (deployed on a cloud server).
[0037] Construct a composite input: Concatenate the preprocessed weighing list image after S1 and the preliminary recognition results output by the small model in S2 into a multimodal input sequence. Input it into the Qwen3-VL-8B multimodal large model (i.e., the multimodal understanding large model), so that on the basis of the visual and text dual channels, it can additionally obtain the guiding signal of "existing recognition suggestions".
[0038] The large model makes comprehensive judgments based on the global layout, context semantics, field logical relationships, etc., and completes three tasks: Correct obvious errors (such as the "6" in the weighing list weight value being misrecognized as "b" by the small model); Complete missing fields (such as content not recognized by the small model); Implement semantic normalization (such as recognizing "total weight = 52.3" as "gross weight").
[0039] Output unified structured JSON format data, and an example is as follows: { "plate_number": "Meng A12345", "weigh_time": "2025-11-05 08:30:00", "gross_weight": 95.6, "tare_weight": 42.3, "net_weight": 33.3, } In an exemplary embodiment, the semantic normalization process specifically includes: Based on a preset field alias dictionary, map similar fields with different naming methods to a unified standard field name.
[0040] Alternatively, based on the numerical relationship between fields, perform semantic inference to map non-standard field names to a unified standard field.
[0041] In an exemplary embodiment, S3 also requires structured information fusion and standardized output, and the executor: the data fusion module (running on a cloud server).
[0042] Integrate the recognition results of the small model and the large model, follow the principle of "the large model dominates and the small model corroborates", and generate a unified structured JSON format data packet, that is, output structured weighing list data, and the field naming specification is as follows: { "plate_number": "Meng A12345", "weigh_time": "2025-11-05 08:30:00", "gross_weight": 45.6, "tare_weight": 12.3, "net_weight": 33.3, } The role of S3 is to provide a consistent data interface for subsequent audits, avoiding the problem of the same thing being called by multiple names.
[0043] In one exemplary embodiment, the logical compliance check includes at least one of the following: Verify whether the net weight in the structured weighing slip data is equal to the difference between the gross weight and the tare weight.
[0044] Verify whether the weighing time in the structured weighing slip data is within the valid business period.
[0045] Verify whether the license plate number format in the structured weighbridge data conforms to national standards.
[0046] Verify whether the weight values in the structured weighbridge data are within a reasonable business range.
[0047] In an exemplary embodiment, the execution entity of S4 is the rule engine module (running on a cloud server).
[0048] Load the pre-configured audit rule library and perform item-by-item verification of the structured data according to the audit rules.
[0049] Output: Each rule returns "pass" or "fail" and records the reason for the exception.
[0050] S4 is used to detect internal data inconsistencies or business violations.
[0051] In one exemplary embodiment, the transport document information includes professional qualification certificate information and / or road transport permit information; the external official authentication interface includes the Ministry of Transport's professional qualification certificate query interface and / or road transport permit query interface provided by the National Government Service Platform.
[0052] In an exemplary embodiment, the execution entity of S5 is an external service docking module (running on a cloud server).
[0053] The authenticity of professional qualification certificates and road transport permits can be verified online by calling the API interfaces provided by the National Government Service Platform (such as the Ministry of Transport's professional qualification certificate query interface and the Ministry of Transport's road transport permit query interface) through the HTTPS protocol.
[0054] The purpose of S5 is to prevent the use of forged or expired documents and enhance the authority of the audit.
[0055] In one exemplary embodiment, S6 specifically includes: S61: When the logic compliance verification result is all passed and the authenticity verification result is true, the conclusion "Audit passed" is generated.
[0056] S62: When a logical verification fails or the document verification result is forged or invalid, generate an "Audit failed" conclusion and record the reason for the exception.
[0057] In an exemplary embodiment, the execution entity of S6 is the audit decision module (running on a cloud server).
[0058] Description: Based on the combined results of S4 (rule verification) and S5 (document verification), the final audit conclusion is generated: Passed: All rules are met and the documents are authentic; Failed: Key information is incorrect or documents are forged.
[0059] In one exemplary embodiment, the application further includes: recording the image processing path, identifying intermediate results, external interface call logs, and audit conclusions to form a traceable audit chain.
[0060] The above steps are executed on a server-side computer system. The execution flow of the computer system is as follows: Figure 2 As shown.
[0061] This application provides an automatic verification and authenticity check system for logistics weighbridge information, including: The image preprocessing module is used to receive the weighbridge image and perform preprocessing to obtain a preprocessed image.
[0062] A dual-layer recognition engine, including: The lightweight OCR recognition submodule is used to perform text recognition on preprocessed images and output preliminary recognition results containing text content, location coordinates, and confidence level.
[0063] The multimodal understanding submodule is used to receive the preprocessed image and the preliminary recognition result, and to correct, complete and semantically normalize the preliminary recognition result based on visual features, text semantics and logical relationships between fields, and output structured bill of lading data.
[0064] The rules engine module is used to load a preset audit rules library and perform logical compliance verification on the structured weighbridge data to determine the logical compliance verification result.
[0065] The external interface proxy module is used to call external official certification interfaces to verify the authenticity of transportation document information and determine the verification result.
[0066] The review decision module is used to generate a final review conclusion by combining the logical compliance verification results and the authenticity verification results.
[0067] In one exemplary embodiment, it further includes: The logging and auditing module is used to record the image processing path, identify intermediate results, external interface call logs, and audit conclusions, forming a traceable audit chain.
[0068] The above modules are deployed on cloud servers.
[0069] This application has the following significant advantages: 1) By using the "small model preprocessing + large model collaborative understanding" mechanism, the recognition accuracy of the large model in the vertical field of weighbridge slips is significantly improved.
[0070] To address the issue of low recognition accuracy of large multimodal models in specialized scenarios such as logistics weighbridge slips (especially for handwritten content), this application creatively introduces a lightweight, dedicated small model as a pre-processor, fully leveraging its high-precision advantage in specific tasks. The small model first performs refined preprocessing on the image, completing recognition and preliminary extraction, generating intermediate results with metadata such as location, content, and confidence level. Subsequently, this result, along with the preprocessed image, is input into the large model, forming an "annotated image-text input." This approach essentially provides the large model with "expert hints," preventing it from understanding the entire image from scratch, but rather allowing it to verify, correct, and complete based on existing recognition suggestions. Experimental data shows that without the guidance of the small model, the large model's recognition accuracy for weighbridge slips is 62.29%; however, after introducing the small model for preprocessing, the recognition accuracy increases to 90.8%. This demonstrates that by providing prior information through the small model, the shortcomings of the large model's insufficient training in vertical domains are effectively compensated, achieving a synergistic gain effect of "1+1>2".
[0071] 2) Small models specialize in handwriting recognition, solving the problem of handwritten weighbridge slip recognition.
[0072] Traditional OCR and general-purpose large models generally suffer from low recognition rates and weak anti-interference capabilities when processing handwritten weighbridge slips. The small model in this application has been trained on a large amount of handwritten weighbridge slip data from real logistics scenarios, and its recognition capabilities for handwritten Arabic numerals, simplified Chinese names, and phrases have been specifically optimized.
[0073] 3) Effectively address the challenges of diverse weighbridge styles and inconsistent field naming, and improve the system's generalization capabilities.
[0074] Faced with the vastly different weighbridge slip templates and diverse field naming conventions across the country, this application ensures recognition stability through a dual mechanism. Firstly, a built-in dynamic field alias dictionary supports the unified mapping of fields such as "Gross Weight," "Gross Weight," "GW," and "Weighing Pre-Weighing Weight" to standard field names. Secondly, the large-scale model learns layout patterns (e.g., the lower right corner is often the signature area, and the left side is often the vehicle information section), enabling it to infer field meanings based on spatial location and semantic relationships even when encountering unfamiliar templates. For example, when a new weighbridge slip uses "Initial Weight" and "Final Weight" to represent weight, the large-scale model can infer that they correspond to "Gross Weight" and "Tare Weight," respectively, based on the relationship that the difference between the two equals the net weight. This combination of "rules + learning" gives the system strong template adaptability and semantic understanding capabilities, eliminating the need to retrain the model for each new format and significantly reducing deployment and maintenance costs.
[0075] 4) Full-process automation and audit traceability mechanisms improve operational efficiency and regulatory transparency.
[0076] This application achieves an end-to-end automated closed loop from image access, information recognition, rule verification, document verification to final decision output. The entire process requires no manual intervention, greatly reducing the workload of operational auditing and other positions. Previously, an auditor could process a maximum of about 300 weighbridge slips per day, and was prone to oversights due to fatigue. This application, however, can operate 24 / 7, with a single server capable of processing over 4,000 slips per day, improving efficiency by more than 70%. Simultaneously, the system is equipped with a complete log recording and audit trail module, meticulously saving every image processing path, recognition result, external interface call record, and final audit conclusion, forming an immutable operational chain. Managers can view statistical reports such as audit pass rates and anomaly type distribution through a visual dashboard to assist in management decisions; regulatory departments can also export all audit records for a specified time period at any time to meet compliance requirements such as tax audits and safety production inspections. This mechanism not only improves internal management efficiency but also enhances the company's risk control capabilities and the credibility of external information disclosure.
[0077] In summary, this application, through its innovative architecture of "small model preprocessing + large model collaborative understanding," successfully addresses industry pain points such as inaccurate recognition of large models in vertical fields, difficulty in recognizing handwritten content, and inconsistent naming conventions. It achieves high-precision, high-efficiency, and highly secure automatic verification and authenticity checking of logistics weighbridge slips. This solution is not only applicable to road freight scenarios but can also be extended to other fields involving weighing and measurement, such as ports, mines, gas stations, and waste collection, demonstrating promising industrialization prospects and patent portfolio value.
[0078] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.
[0079] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0080] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.
Claims
1. A method for automatically verifying and confirming the authenticity of logistics weighbridge information, characterized in that, include: Receive the weighbridge image and preprocess it to obtain a preprocessed image; The preprocessed image is input into a lightweight OCR recognition model for text recognition, and the output is a preliminary recognition result containing text content, location coordinates and confidence level. The preprocessed image and the preliminary recognition result are jointly input into the multimodal understanding model. Based on visual features, text semantics and logical relationships between fields, the preliminary recognition result is corrected, completed and semantically normalized, and structured bill of lading data is output. The structured weighbridge data is logically validated based on a pre-defined audit rule base to determine the logical compliance validation result. The external official authentication interface is called to verify the authenticity of the transportation document information contained in the structured weighbridge data, and the verification result is determined. The final audit conclusion is generated by combining the logical compliance verification results with the authenticity verification results.
2. The method for automatic verification and authenticity checking of logistics weighbridge information according to claim 1, characterized in that, The preprocessed image and the preliminary recognition result are jointly input into the multimodal understanding model, specifically including: The preliminary recognition results are superimposed onto the preprocessed image in the form of annotations to form an annotated composite input image; The composite input image is input into the multimodal understanding model, which then guides the model to perform contextual reasoning and result optimization based on the annotations.
3. The method for automatic verification and authenticity checking of logistics weighbridge information according to claim 1 or 2, characterized in that, The lightweight OCR recognition model is fine-tuned using a dedicated training dataset. The dedicated training dataset contains handwritten weighing slip samples from logistics scenarios, which enhances the recognition capability of the lightweight OCR recognition model for handwritten Arabic numerals and Chinese characters.
4. The method for automatic verification and authenticity checking of logistics weighbridge information according to claim 1, characterized in that, The semantic normalization process specifically includes: Based on a pre-defined field alias dictionary, fields of the same type with different naming methods are mapped to a unified standard field name; or, Semantic inference is performed based on the numerical relationships between fields to map non-standard field names to unified standard fields.
5. The method for automatic verification and authenticity checking of logistics weighbridge information according to claim 1, characterized in that, Logical compliance checks include at least one of the following: Verify whether the net weight in the structured weighing slip data is equal to the difference between the gross weight and the tare weight; Verify whether the weighing time in the structured weighing slip data is within the valid business period; Verify whether the license plate number format in the structured weighbridge data conforms to national standards; Verify whether the weight values in the structured weighbridge data are within a reasonable business range.
6. The method for automatic verification and authenticity checking of logistics weighbridge information according to claim 1, characterized in that, The transportation document information includes professional qualification certificate information and / or road transport permit information; the external official authentication interface includes the Ministry of Transport's professional qualification certificate query interface and / or road transport permit query interface provided by the National Government Service Platform.
7. The method for automatic verification and authenticity checking of logistics weighbridge information according to claim 1, characterized in that, Based on the combined results of the logical compliance verification and the authenticity verification, a final audit conclusion is generated, which specifically includes: When the logical compliance verification result is all passed and the authenticity verification result is true, a "Review passed" conclusion is generated; When a logical check fails or the document verification result is found to be forged or invalid, a "Review Failed" conclusion is generated and the reason for the exception is recorded.
8. The method for automatic verification and authenticity checking of logistics weighbridge information according to claim 1, characterized in that, Also includes: Record the image processing path, identify intermediate results, external interface call logs, and audit conclusions to form a traceable audit chain.
9. An automatic verification and authenticity check system for logistics weighbridge information, characterized in that, The automatic verification and authenticity check system for logistics weighbridge information executes the automatic verification and authenticity check method for logistics weighbridge information according to any one of claims 1-8, wherein the automatic verification and authenticity check system for logistics weighbridge information includes: The image preprocessing module is used to receive the weighbridge image and preprocess it to obtain a preprocessed image; A dual-layer recognition engine, including: A lightweight OCR recognition submodule is used to perform text recognition on preprocessed images and output preliminary recognition results containing text content, location coordinates, and confidence level. The multimodal understanding submodule is used to receive the preprocessed image and the preliminary recognition result, and to correct, complete and semantically normalize the preliminary recognition result based on visual features, text semantics and logical relationships between fields, and output structured bill of lading data. The rules engine module is used to load a preset audit rules library and perform logical compliance verification on the structured weighbridge data to determine the logical compliance verification result. The external interface proxy module is used to call external official certification interfaces to verify the authenticity of transportation documents and determine the verification result. The review decision module is used to generate a final review conclusion by combining the logical compliance verification results and the authenticity verification results.
10. The automatic verification and authenticity check system for logistics weighbridge information according to claim 9, characterized in that, Also includes: The logging and auditing module is used to record the image processing path, identify intermediate results, external interface call logs, and audit conclusions, forming a traceable audit chain.