Vehicle work order verification method and device

Abstract

The invention discloses a vehicle work order verification method and device, and relates to the technical field of computers. A specific embodiment of the method comprises the steps of obtaining a work order picture detection result and a work order text extraction result corresponding to a to-be-verified vehicle work order; carrying out alias conversion processing on the work order text extraction result based on a pre-configured alias library to obtain a converted work order text extraction result; performing data consistency verification based on the work order picture detection result and the converted work order text extraction result to obtain an image-text data verification result; and extracting image features and text features from the image-text data verification result, and determining a reexamination verification result of the to-be-verified vehicle work order based on the image features and the text features. According to the embodiment of the invention, work order verification can be automatically carried out, the verification cost is reduced, and the verification efficiency and the accuracy of the verification result are improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method and device for verifying a vehicle work order. Background Art

[0002] The current vehicle work order verification method mainly involves manually listing the verification content, then conducting a simple preliminary review based on the verification rules, and finally having the back-end staff manually perform subsequent verification processing.

[0003] In the process of implementing the present invention, the inventors discovered that there are at least the following problems in the related art:

[0004] Work order verification cannot be performed automatically, which results in high costs, low efficiency, and poor accuracy of verification results. Summary of the Invention

[0005] In view of this, an embodiment of the present invention provides a method and device for verifying a vehicle work order, which can automatically verify the work order, reduce the verification cost, and improve the verification efficiency and the accuracy of the verification results.

[0006] To achieve the above object, according to one aspect of an embodiment of the present invention, a method for verifying a vehicle work order is provided, comprising:

[0007] Obtain the work order image detection results and work order text extraction results corresponding to the vehicle work order to be verified;

[0008] Perform alias conversion processing on the work order text extraction result based on the pre-configured alias library to obtain the converted work order text extraction result;

[0009] Perform data consistency verification based on the work order image detection results and the converted work order text extraction results to obtain the image and text data verification results;

[0010] Image features and text features are extracted from the image and text data verification results, and the review verification results of the vehicle work order to be verified are determined based on the image features and text features.

[0011] Optionally, the method further comprises:

[0012] The compatibility of the vehicle and the product in the vehicle work order to be verified is verified according to the pre-configured adaptation relationship, and the adaptation relationship verification result of the vehicle work order to be verified is obtained.

[0013] Optionally, determining a review and verification result of the vehicle work order to be verified based on image features and text features includes:

[0014] Perform weighted fusion of image features and text features to obtain an image-text fusion vector;

[0015] Based on the image-text fusion vector and the acquired historical image-text fusion vector, the work order image data and work order text data corresponding to the vehicle work order to be verified are re-verified to obtain a review verification result of the vehicle work order to be verified.

[0016] Optionally, the method further comprises:

[0017] Determine a price verification strategy based on the acquired historical vehicle work order data; verify the amount data in the vehicle work order to be verified according to the price verification strategy to obtain a price risk verification result for the vehicle work order to be verified; wherein the historical vehicle work order data refers to historical data of work orders of the same category or similar categories as the vehicle work order to be verified;

[0018] Alternatively, an address verification strategy is determined based on the acquired historical vehicle work order data; the regional data in the vehicle work order to be verified is verified according to the address verification strategy to obtain an address risk verification result of the vehicle work order to be verified.

[0019] Optionally, before obtaining the inspection result of the work order image corresponding to the vehicle work order to be verified, the following steps are also included:

[0020] Using the work order image detection model, the obtained work order image data corresponding to the vehicle work order to be verified is detected and processed to obtain a work order image detection result corresponding to the vehicle work order to be verified; wherein the work order image detection result includes at least one of the following: a vehicle body image detection result, a vehicle model image detection result, and a license plate image detection result;

[0021] The method also includes:

[0022] Iteratively optimize the work order image detection model based on the work order image detection results.

[0023] Optionally, before obtaining the work order text extraction result corresponding to the vehicle work order to be verified, the following steps are further included:

[0024] Performing entity extraction processing on the acquired work order text data corresponding to the work order for the vehicle to be verified using the work order entity extraction model to obtain a work order text extraction result corresponding to the work order for the vehicle to be verified; wherein the work order text extraction result includes at least one of the following: a vehicle model entity extraction result and a license plate entity extraction result;

[0025] The method also includes:

[0026] The work order entity extraction model is iteratively optimized based on the work order text extraction results.

[0027] According to a second aspect of an embodiment of the present invention, a vehicle work order verification device is provided, comprising:

[0028] The result acquisition module is used to obtain the work order image detection results and work order text extraction results corresponding to the vehicle work order to be verified;

[0029] The alias conversion module is used to perform alias conversion processing on the work order text extraction result based on the pre-configured alias library to obtain the work order text extraction result after conversion;

[0030] The data verification module is used to perform data consistency verification based on the work order image detection results and the work order text extraction results after conversion processing to obtain the image and text data verification results;

[0031] The review and verification module is used to extract image features and text features from the image and text data verification results, and determine the review and verification results of the vehicle work order to be verified based on the image features and text features.

[0032] According to a third aspect of an embodiment of the present invention, there is provided an electronic device, including:

[0033] one or more processors;

[0034] a memory for storing one or more programs,

[0035] When one or more programs are executed by one or more processors, the one or more processors are enabled to implement the method of any of the above embodiments.

[0036] According to a fourth aspect of an embodiment of the present invention, a computer-readable medium is provided, on which a computer program is stored. When the program is executed by a processor, the method of any of the above embodiments is implemented.

[0037] According to a fifth aspect of an embodiment of the present invention, a computer program product is provided, including a computer program, which implements the method of any of the above embodiments when executed by a processor.

[0038] One embodiment of the above invention has the following advantages or beneficial effects: by obtaining the work order image detection result and the work order text extraction result corresponding to the vehicle work order to be verified; performing alias conversion processing on the work order text extraction result based on a pre-configured alias library to obtain the work order text extraction result after the conversion processing; performing data consistency verification based on the work order image detection result and the work order text extraction result after the conversion processing to obtain the image and text data verification result; extracting image features and text features from the image and text data verification result, and determining the review verification result of the vehicle work order to be verified based on the image features and text features; thereby, the work order verification can be automatically performed, and the verification cost is reduced, and the verification efficiency and the accuracy of the verification results are improved.

[0039] The further effects of the above-mentioned non-conventional optional manner will be described below in conjunction with specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The accompanying drawings are provided for a better understanding of the present invention and are not intended to limit the present invention.

[0041] Figure 1 1 is a schematic diagram of the main process of a method for verifying a vehicle work order according to an embodiment of the present invention;

[0042] Figure 2 1 is a schematic diagram of the main process of a method for verifying a vehicle work order according to a reference embodiment of the present invention;

[0043] Figure 3 1 is a schematic diagram of the main process of a method for verifying a vehicle work order according to a preferred embodiment of the present invention;

[0044] Figure 4 is a schematic diagram of main modules of a vehicle work order verification device according to an embodiment of the present invention;

[0045] Figure 5 is an exemplary system architecture diagram in which embodiments of the present invention may be applied;

[0046] Figure 6 It is a schematic diagram of the structure of a computer system of a terminal device or a server suitable for implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The following description of exemplary embodiments of the present invention is made in conjunction with the accompanying drawings, in which various details of the embodiments of the present invention are included to facilitate understanding. These details should be considered as merely exemplary. Therefore, it should be appreciated by those skilled in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0048] It should be noted that the acquisition, storage and application of personal information involved in the embodiments of the present invention are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0049] At present, the verification contents are mainly listed manually, and then a simple preliminary review is performed according to the verification rules. Finally, the back-end staff manually performs subsequent verification processing. Work order verification cannot be performed automatically, which is costly, inefficient, and the accuracy of the verification results is poor.

[0050] In view of this, according to one aspect of an embodiment of the present invention, a method for verifying a vehicle work order is provided.

[0051] Figure 1 FIG. 1 is a schematic diagram of the main process of the vehicle work order verification method according to an embodiment of the present invention. Figure 1As shown, the vehicle work order verification method according to an embodiment of the present invention includes the following steps S101 to S104.

[0052] Step S101: Obtain the work order image detection result and work order text extraction result corresponding to the vehicle work order to be verified.

[0053] Among them, the vehicle work order to be verified refers to the vehicle work order information that has not yet been confirmed or reviewed, and may include information such as the work order number, creation time, responsible person, and work order content description. The vehicle work order to be verified can come from different sources, such as the product verification form provided offline by the user, the repair request submitted by the user, the internally generated maintenance task, or other forms of work order records. The work order image detection results include information such as the detected object category, location coordinates (such as a bounding box), and confidence score. For example, if the work order image data involves a failure of a certain device, the work order image detection model can mark the damaged parts on the device and provide detailed information about the damage and the possible type of failure. The work order text extraction results include key information extracted from the work order description or related documents, such as the failure description, work content, technical specifications, etc.

[0054] Specifically, a pre-trained image detection model can be used to extract information from vehicle work order images to generate work order image detection results. Alternatively, methods based on edge detection or segmentation algorithms can be used to locate key areas within work order images, and rule matching can be used to detect specific object categories and locations to generate work order image detection results. Regarding work order text extraction results, a work order entity extraction model can be used to perform entity extraction on work order text data to generate work order text extraction results. Alternatively, text regions within work order images can be identified and combined with natural language processing techniques to extract key fields and perform semantic analysis to generate work order text extraction results.

[0055] Step S102 : performing alias conversion processing on the work order text extraction result based on a pre-configured alias library to obtain a work order text extraction result after conversion processing.

[0056] Using a pre-configured alias library, different forms of entity names are uniformly converted into standardized names, thereby generating more accurate and consistent work order text extraction results. The entity data extracted from the work order text data may contain various naming forms. For example, the same brand, model, or part may have multiple aliases or abbreviations. To ensure data consistency, the system loads the pre-configured alias library, which stores various aliases of common entities and their corresponding standardized names. For example, "Toyota" and "丰田" are unified as "Toyota", or "发动机" and "引擎" are unified as "发动机", etc. Perform alias conversion processing on the extracted work order text extraction results. Specifically, scan the entities in the extracted work order text extraction results one by one and look up their corresponding standardized names in the alias library. If a match is found, the system will replace the original entity name with the standardized name. For example, if the entity extracted from the work order text data is "丰田", the system will convert it to the standardized name "Toyota". This conversion is not limited to brand names but also includes other entity types, such as part names, model names, etc. Through alias conversion processing, data confusion and errors caused by different naming methods are avoided, and more accurate and consistent work order text extraction results can be generated, providing a solid data foundation for subsequent verification and processing, and improving the efficiency and accuracy of work order processing.

[0057] Step S103, perform data consistency verification based on the work order picture detection result and the work order text extraction result after conversion processing to obtain the graphic and text data verification result.

[0058] Data consistency verification is performed based on the work order image detection results and the converted work order text extraction results, ensuring the integrity and consistency of the work order data to be verified in both image and text dimensions. Specifically, structured data is obtained from the work order image detection results and the converted work order text extraction results. The work order image detection results include key information extracted from the image, such as object detection (e.g., equipment, components), region identification (e.g., location coordinates), and specific attributes (e.g., damage level or status). The work order text extraction results include key information extracted from the work order description or related documents, such as the fault description, work content, and technical specifications. Data matching and verification are then performed to compare the objects, regions, or attributes detected in the work order image with the descriptions and information extracted from the work order text. For example, this verifies whether the specific equipment or problem described in the work order matches the actual information detected in the image. If inconsistencies or incompleteness are found between the work order image and text, the image or text data can be retrieved to further verify the specific content of the work order to improve the accuracy and credibility of the final verification results. Finally, the image and text data verification results for the work order to be verified are generated. These results summarize the consistency assessment of each piece of data to be verified, reflecting the comprehensive performance of the image and text data in terms of matching and accuracy. These verification results provide important basis for subsequent approval, review, or decision-making, ensuring the reliability and effectiveness of the work order information processing process.

[0059] Step S104 , extracting image features and text features from the image and text data verification results, and determining a review verification result of the vehicle work order to be verified based on the image features and text features.

[0060] Image features and text features are extracted from the image and text data verification results. Image features can include visual information such as the object category, location, color, shape, etc. detected in the image, which can be extracted through deep learning models (such as convolutional neural networks) or other methods; text features include key information in the work order text, such as equipment name, fault description, operation steps, etc., which can be extracted through natural language processing technology (such as the BERT model) or other methods.

[0061] The embodiment of the present invention obtains the work order image detection result and the work order text extraction result corresponding to the vehicle work order to be verified; performs alias conversion processing on the work order text extraction result based on a pre-configured alias library to obtain the work order text extraction result after the conversion processing; performs data consistency verification based on the work order image detection result and the work order text extraction result after the conversion processing to obtain the image and text data verification result; extracts image features and text features from the image and text data verification result, and determines the review verification result of the vehicle work order to be verified based on the image features and text features; it can automatically perform work order verification, reduce verification costs, and improve verification efficiency and the accuracy of verification results.

[0062] Optionally, the method further includes: verifying the compatibility of the vehicle and the commodity in the vehicle work order to be verified according to a pre-configured adaptation relationship, and obtaining an adaptation relationship verification result of the vehicle work order to be verified.

[0063] Vehicle and product information is extracted from the vehicle work order to be verified. Vehicle information includes key attributes such as brand, model, and model year, while product information includes product type, specifications, and applicable scope. Pre-configured compatibility relationships are stored in a relationship database, which stores the vehicle-product correspondence and compatibility rules. These rules can be developed based on historical data, expert experience, and actual usage. Using the pre-configured compatibility relationship database, the extracted vehicle information is matched and compared with the product information. Specifically, the system searches the compatibility relationship database to find product compatibility rules that match the current vehicle information. For example, if the vehicle in the work order to be verified is a specific model of a certain brand, the system can retrieve a detailed list of compatible consumables, accessories, and other products for that model, along with their compatibility conditions. During this comparison process, the system determines whether the product in the work order meets the vehicle's compatibility conditions. This includes checking whether the product specifications match the vehicle's technical requirements, such as size, model, and compatibility. Furthermore, the system verifies whether the product's applicable scope covers the current vehicle's brand and model. If multiple products meet the compatibility criteria, the matching criteria can be further refined to select the most suitable product for comparison. If any inconsistencies or non-compliance with the compatibility criteria are discovered during the matching process, the pending verification work order is marked as unqualified for further verification and processing, with product or vehicle information reconfirmed or the content of the pending vehicle work order adjusted to meet the compatibility requirements. Finally, the compatibility verification results for the pending vehicle work order are generated. These results include an assessment of the compatibility of the vehicle and product, a detailed explanation of whether the compatibility criteria are met, any reasons for non-compliance, and a recommended solution.

[0064] The embodiment of the present invention verifies the matching degree between the commodity and the vehicle in the vehicle work order to be verified through the adaptation relationship, improves the accuracy and reliability of vehicle work order processing, and enhances the quality and efficiency of overall business operations.

[0065] Optionally, Figure 2 FIG. 1 is a schematic diagram of the main process of a method for verifying a vehicle work order according to a reference embodiment of the present invention. Figure 2 As shown, the review and verification result of the vehicle work order to be verified is determined based on the image features and text features, including steps S201 to S202.

[0066] Step S201: Perform weighted fusion on the image features and text features to obtain an image-text fusion vector.

[0067] Step S202 : Based on the image-text fusion vector and the acquired historical image-text fusion vector, re-verify the work order image data and work order text data corresponding to the vehicle work order to be verified, and obtain a review verification result of the vehicle work order to be verified.

[0068] The extracted image and text features are weighted and fused to produce a fused image-text vector. Weighted fusion assigns weights based on the importance and relevance of image and text features, enabling the fused vector to more comprehensively represent the overall information of the work order. Fusion methods can employ various techniques, such as weighted averaging and attention mechanisms, to ensure that the fused vector fully captures the relevance and consistency of the image and text data.

[0069] Based on the generated image-text fusion vector and the acquired historical image-text fusion vector, the work order image and text data corresponding to the vehicle work order to be verified are re-verified. The historical image-text fusion vector stores the image and text feature information of previous work orders. By comparing it with the image-text fusion vector of the current work order to be verified, duplicate content, anomalies, and potential issues can be identified.

[0070] During the review and verification process, the system compares the image-text fusion vectors of the current vehicle work order to be verified with those of historical vehicle work orders, checking for duplicate submissions or similar issues to ensure the uniqueness and accuracy of the current vehicle work order information. By analyzing the image-text fusion vectors, anomalies in the text and image data, such as abnormal amounts or unusual locations, are identified and flagged for resolution. This comprehensive analysis of fusion vectors improves the accuracy of vehicle work order reviews and reduces misjudgments caused by a single data source.

[0071] Optionally, the method also includes: determining a price verification strategy based on the acquired historical vehicle work order data; verifying the amount data in the vehicle work order to be verified according to the price verification strategy, and obtaining a price risk verification result of the vehicle work order to be verified; wherein the historical vehicle work order data is historical data of work orders of the same category or similar categories of the vehicle work order to be verified; or, determining an address verification strategy based on the acquired historical vehicle work order data; verifying the regional data in the vehicle work order to be verified according to the address verification strategy, and obtaining an address risk verification result of the vehicle work order to be verified.

[0072] Obtain work order data of the same or similar categories from historical vehicle work orders. Historical vehicle work orders include price information, address information, and other relevant attributes. By analyzing these work order data, the price range and address distribution under normal circumstances can be determined. Among them, work orders of the same category refer to work orders belonging to the same category or classification. These work orders have a high degree of consistency in attributes, processing procedures, and business content. For example, the work orders involve the same type of goods or services, and are all about car repair work orders of a certain brand and model, or after-sales service work orders for a certain product. Work orders of similar categories refer to work orders that have certain similarities with the work order to be verified in terms of attributes, processing procedures, and business content, but are not exactly the same. For example, similar categories of goods or services, car repair work orders of the same brand but different models, or after-sales work orders for different types of electronic products.

[0073] Based on historical vehicle work orders, the system analyzes the price distribution of similar or similar work orders to determine a reasonable price range and criteria for identifying abnormal prices. For example, through statistical analysis, the system calculates the mean and standard deviation of normal prices to determine upper and lower price limits. Based on a defined price verification strategy, the system verifies the amount in the work order data for verification. If the amount in the work order for verification exceeds the predetermined reasonable price range, the system flags it as a price risk anomaly and generates a price risk verification result, which can identify potential price fraud or data entry errors. By analyzing the address data of historical work orders, the system identifies common geographic distributions and abnormal areas. Historical data analysis can include methods such as frequency analysis and geographic clustering to identify normal service areas and uncommon abnormal addresses. Based on a defined address verification strategy, the system verifies the address information in the work order data for verification. The system checks whether the address of the work order for verification is within the normal service area, whether there are unreasonable geographical locations, or whether the address changes frequently. If an abnormal address is found, it is flagged as an address risk anomaly and generates an address risk verification result, which helps identify potential address fraud or operational errors.

[0074] Optionally, before obtaining the work order image detection result corresponding to the vehicle work order to be verified, the method further includes: using a work order image detection model to detect and process the obtained work order image data corresponding to the vehicle work order to be verified to obtain the work order image detection result corresponding to the vehicle work order to be verified; wherein the work order image detection result includes at least one of the following: a vehicle body image detection result, a vehicle model image detection result, and a license plate image detection result. The method further includes: iteratively optimizing the work order image detection model based on the work order image detection result.

[0075] Work order image data refers to images or graphic materials related to the work order. Images can be on-site photos, equipment screenshots, fault diagrams, or other relevant visual materials. Images are uploaded together with the work order or collected through other means, and are associated with information such as the work order number and stored in the database. You can use the query function or API (Application Programming Interface) interface of the work order management system to obtain a list of all work orders in a pending verification state, or scan the work orders to be verified provided by the user offline, and extract basic information about the work orders to be verified from the database or system, such as the work order number, creation time, responsible person, and work order status. For each work order to be verified, the corresponding image file is retrieved from the image storage system (such as a file server, cloud storage, etc.) using the work order number. You can use the API interface, file path, or other associated information to obtain relevant images and match the images with the correct work order.

[0076] Specifically, this embodiment inputs work order-related image data into a pre-trained work order image detection model to identify and analyze key information within the image. The work order image detection model is capable of recognizing specific types of image features and anomalies. It performs multi-level feature extraction and analysis on the work order image data, identifying important elements within the image, such as equipment components, equipment appearance, damaged areas, operating tools, and so on, and annotating the detected objects or anomalies. After the detection process is complete, the work order image detection model generates work order image detection results. This information can be cross-validated with the text description of the work order, improving the accuracy and efficiency of work order verification.

[0077] In addition, the work order image detection results are collected and organized. The work order image detection results come from the actual work order processing process and include the work order image detection output for the vehicle body, vehicle model, and license plate. These detection results include the model's predicted label, prediction confidence, the position and size of the detection box, and the comparison results with the actual situation. For the vehicle body image detection results, the work order image detection model identifies and locates the outline and features of the car. During the iterative optimization process, cases of misclassification or inaccurate positioning in the work order image detection results are analyzed. For example, the model may mistakenly identify other objects as vehicle bodies or fail to correctly select the vehicle body area. Data enhancement is performed to address weaknesses in the detection. For example, training data is enriched by adding vehicle body images from different angles and under different lighting conditions. Based on the error analysis results, the model architecture or parameters are adjusted, such as adjusting the depth of the convolution layer, adding feature extraction layers, etc., to improve the model's accuracy in vehicle body detection. For vehicle image detection results, the work order image detection model accurately identifies the make and model of the car. During the iterative optimization process, the vehicle model is subdivided into more specific categories, and the precision of the labels is increased to improve the model's ability to distinguish similar models. The imbalance in the number of samples of different vehicle models in the training data is addressed by resampling the data or generating synthetic data to balance the sample distribution. Feature extraction methods are improved, utilizing multi-scale features or fusing multiple feature extraction techniques to increase the model's sensitivity to vehicle model details. For license plate image detection results, the work order image detection model accurately locates and recognizes license plate characters. During the iterative optimization process, character recognition errors are analyzed, the character recognition algorithm is improved, and a more powerful OCR (optical character recognition) model is adopted. Noise and blur in license plate images are addressed, such as by using image preprocessing techniques to remove noise and enhance contrast, thereby improving the clarity of license plate characters. Transfer learning or few-shot learning techniques are used to improve the model's performance on small sample data, especially for the recognition of rare license plate styles or special characters.

[0078] Through the above optimization steps, the work order image detection model in the embodiment of the present invention can be more accurate and reliable in detecting vehicle body, vehicle model and license plate, thereby improving the efficiency and quality of overall work order processing and enhancing user satisfaction.

[0079] Optionally, before obtaining the work order text extraction result corresponding to the vehicle work order to be verified, the method further includes: performing entity extraction processing on the obtained work order text data corresponding to the vehicle work order to be verified using a work order entity extraction model to obtain a work order text extraction result corresponding to the vehicle work order to be verified; wherein the work order text extraction result includes at least one of the following: a vehicle model entity extraction result and a license plate entity extraction result. The method further includes iteratively optimizing the work order entity extraction model based on the work order text extraction result.

[0080] Work order text data refers to the detailed text description of a work order, including but not limited to the fault description, solution, operation steps, and customer feedback. This work order text data can be stored in the work order management system's database and retrieved and linked using identifiers such as the work order number. Work order text data can be obtained by extracting the detailed text description of the work order based on the work order number through the work order management system's database query function or API interface.

[0081] Specifically, natural language processing techniques are used to automatically extract key entity information from work order text data to support detailed analysis and verification of work order content. The work order entity extraction model, based on machine learning algorithms such as sequence labeling models (e.g., BiLSTM-CRF) or pre-trained language models (e.g., BERT), learns to identify and classify entities in text, such as license plate numbers, vehicle models, and brands, by training on large amounts of labeled data. The work order entity extraction model performs multi-level text analysis and semantic understanding on work order text data, identifying and annotating the location and type of specific entities within the work order text. For example, it can extract important elements such as the vehicle's license plate number, brand information, and vehicle model information from the work order text data. The output of the work order entity extraction model can be structured data, including the extracted entity type (e.g., license plate, brand, and vehicle model), the entity's specific content, and the entity's location within the text. This data not only helps identify key information within the work order content but also provides a foundation for subsequent automated processing. The work order entity extraction model can significantly improve the efficiency and accuracy of work order processing. Compared with traditional manual processing, this automated text analysis technology can quickly and accurately extract important information from work orders, reducing the time and cost of manual processing while also reducing problems caused by human error. It is not only suitable for processing a single work order, but can also effectively meet the needs of efficient processing of large amounts of work order data, providing enterprises with more efficient and intelligent business operations support.

[0082] By analyzing errors and deficiencies in the work order text extraction results, we can identify issues with the work order entity extraction model in extracting vehicle model and license plate entities. For example, the work order entity extraction model may incorrectly extract irrelevant information or inaccurately distinguish similar vehicle models and license plates. To optimize vehicle model entity extraction results, we can increase the diversity and coverage of training data. Especially for easily confused vehicle model entities, expanding the dataset to include more text data from different brands, models, and years can enhance the model's recognition capabilities for various vehicle types. Furthermore, we can employ more advanced natural language processing techniques, such as context-aware language models (e.g., BERT), to improve the model's understanding and extraction of vehicle model entities. Regarding license plate entity extraction results, license plate entities typically have a fixed format and character combination. A rule-based model-based correction mechanism can be introduced to verify and correct the extracted license plate entities. Furthermore, by training on license plate data from different regions and types, we can enhance the generalization and recognition accuracy of the work order entity extraction model. During the entire iterative optimization process, we should not only focus on the effect of a single optimization, but also establish a continuous optimization mechanism. By regularly collecting new work order text data, we can continuously update and expand the training set to maintain the model's efficient learning and adaptability.

[0083] The embodiment of the present invention optimizes the work order entity extraction model based on work order text extraction, which can make the work order text extraction model more accurate and reliable in extracting vehicle model entities and license plate entities, thereby improving the overall efficiency and accuracy of work order processing.

[0084] Figure 3 FIG. 1 is a schematic diagram of the main process of the vehicle work order verification method according to the preferred embodiment of the present invention. Figure 3 As shown, this embodiment is a scenario of offline verification of the correctness of the installation of automobile wearing parts, which mainly includes the following steps 1 to 8.

[0085] Step 1: Input the work order data to be verified, and load the product attributes, work order text, and work order image information from the product attribute library, work order text library, and work order image library.

[0086] Step 2: For the work order image included in the picture, it is necessary to complete the vehicle body detection, license plate detection, and car brand and model recognition in the work order image. The task involves multi-target task detection. It is necessary to use real high-quality images to fine-tune the model to improve the accuracy of vehicle body detection, license plate detection, and car brand recognition, and complete the optimization and training of the image extraction model.

[0087] Step 3: For the work order image data, use the work order image detection model trained in step 2 to complete the vehicle body detection, brand and model recognition, and license plate detection model work. After license plate detection, the actual license plate needs to be identified through the OCR recognition model optimized for license plates to complete the information extraction contained in the car work order image.

[0088] Step 4: For the work order text data, use the work order text entity extraction model trained based on the automobile work order text data to extract the license plate information, brand, and model information from the automobile work order text.

[0089] Step 5: In steps 3 and 4, the vehicle body detection, vehicle model recognition, license plate detection, and license plate recognition work in the work order image have been completed, and the license plate entity extraction and vehicle model entity extraction work in the work order text have been completed, and the basic text data extracted by the original model has been obtained. Next, it is necessary to verify the consistency of the image and text data, mainly to verify the license plate consistency, vehicle model consistency, and whether the vehicle body threshold of the verification work order meets the work order verification conditions. During the verification process, the non-normalized entities extracted from the text of the verification work order need to be normalized into entities consistent with the image dimensions, and then compared.

[0090] Step 6: Through the first five steps, it can be basically verified whether the vehicle recorded in the work order matches the vehicle that actually arrived at the store. It is necessary to further verify whether the compatibility relationship between the product and the vehicle model meets the conditions. Different products have different verification attributes. The most similar product and the vehicle model compatibility relationship of the corresponding product are obtained. Then, it is verified whether the vehicle model of the work order is compatible with the current product.

[0091] Step 7: After completing the verification of a single sales and proofreading work order in the first six steps, the backbone network of the image extraction network and the penultimate layer of the named entity extraction model are used to complete the feature extraction of images and texts. After unifying the feature dimensions, they are fused into an image-text fusion vector using a weighted average method. The image-text vector recall engine is then used to verify whether the current work order image, text, and historical work order images and texts have duplicate uploaded content. It is also necessary to verify whether the current work order has any abnormal amounts or common regional anomalies in historical work orders of the same category. This is used to add verification logic for historical data dimensions and complete retrospective review work.

[0092] Step 8: After completing the above 7 steps, the final multimodal review and verification results of the verification work order are obtained, and the incremental data is stored to verify the compliance results of the verification work order and output to downstream applications.

[0093] Specifically, from system input to output, an embodiment of the present invention is described in detail according to seven key modules: work order verification image information extraction module, verification text entity extraction module, image and text consistency verification module, commodity adaptation relationship verification module, historical image and text duplicate retrieval verification module, and historical work order anomaly detection and storage output.

[0094] The verification image information extraction module uses real-life photos of technicians performing on-site construction and verification of product replacement. Its main goal is to detect whether a vehicle is included in the verification work order, whether the vehicle body ratio meets the photo threshold, the specific model name of the photographed vehicle, and the license plate of the photographed vehicle. After the detection is completed, the detected license plate needs to be used to identify the license plate text using an OCR model (Optical Character Recognition). Therefore, it is necessary to complete vehicle multi-information detection, OCR recognition model fine-tuning, vehicle information recognition, and threshold calculation.

[0095] Vehicle multi-information detection and recognition and OCR recognition model fine-tuning first involves image data collection and preprocessing, mainly through stratified sampling and screening of image data from historical work orders; acquisition of actual vehicle construction images: based on the existing work order images, collect images of all brands of vehicles on the market, including images of vehicles of different brands, models, colors and construction environments; acquisition of license plate images: collect license plate images to ensure that license plates from different regions, fonts and formats are included; data annotation: annotate vehicle images, including whether they are cars, vehicle brands, vehicle models, etc., and perform OCR annotation on license plate images; data enhancement: increase data diversity through methods such as rotation, scaling, and cropping; data cleaning: remove poor quality images, and at the same time ensure brand coverage, vehicle model coverage, and license plate character coverage to ensure the quality of the data set. Then comes the work of model fine-tuning and iterative model deployment; Model selection: For the vehicle detection model, select a pre-trained target recognition and detection model (such as YOLO, SSD, Faster R-CNN) for fine-tuning; for license plate OCR recognition, select a pre-trained OCR model (such as CRNN, Attention OCR) for fine-tuning. Before fine-tuning, the data format needs to be converted and adjusted to the corresponding model to adapt to the model input; Vehicle identification and detection model fine-tuning: Use the vehicle image dataset to fine-tune the model, adjust the last few layers of the model to adapt to the new categories (vehicle brand, vehicle model) and accurately draw the object detection box, and train the model until a satisfactory accuracy rate is achieved on the validation set; OCR recognition model fine-tuning: Use the license plate image dataset to fine-tune the model to adapt it to the recognition results of the current license plate character type, and train the model until the business recognition rate is achieved on the validation set;

[0096] Vehicle information recognition OCR recognition and threshold calculation. The above-mentioned vehicle multi-information detection and recognition model and OCR recognition model have been fine-tuned using real verification work order images (vehicle images and license plate images). The trained model is exported to a deployable format (such as ONNX, TensorRT, etc.) and encapsulated as an API service for easy integration into the entire application process. The model is optimized according to the actual application scenario to ensure that it meets the requirements of real-time and batch processing. At the same time, due to the continuous updates and iterations of the vehicle market, the model also needs to be iterated with the update of new data. Therefore, it is necessary to collect model prediction results and user feedback in actual applications, update the dataset based on the feedback, including adding new samples and correcting incorrect annotations, and regularly retrain and fine-tune the model using the updated dataset, and automatically update the hotline model. When a new verification work order image is processed by the vehicle multi-information detection and recognition model and the OCR recognition model, the label L indicating whether it is a car, the coordinates of the vehicle model detection box after identification as a car, Loc1 = [x1, y1, x2, y2], and the standardized vehicle brand can be obtained. 、Model 、Year , license plate The length, width, and H dimensions of the original work order image are used to calculate the ratio of the vehicle body to the total image. This is used to assess whether the ratio meets the work order verification body ratio requirements.

[0097]

[0098] The verification text entity extraction module verifies the personnel information entered in the verification work order after the verification of the verification image is completed. The strategy uses a fine-tuned named entity extraction model to perform text verification and extraction, extracting information such as the vehicle brand, vehicle model, and entered license plate number from the text. At the same time, the extracted names are converted to aliases for matching the vehicle model information results of the image recognition in the above steps.

[0099] The fine-tuning of the named entity model begins with the collection and preprocessing of text data, which is mainly done through stratified sampling and screening of text data from historical work orders; Work order text data acquisition: Use historical work order data for input, collect text data containing vehicle brands and model names, which need to include entities such as model brands and models’ standard names, aliases, nicknames, and year models; Data annotation and preprocessing: Manually annotate the collected text data, mark out entities such as vehicle brands and model names in the text using the BIO annotation specification, and then perform steps such as text cleaning (removing irrelevant characters, unifying character encoding, etc.), word segmentation, and vocabulary construction on the text. At the same time, through the reorganization of entity positions and alias replacement, Based on the original data, the data scale and training data are further expanded to ensure the quality of text data; Model selection: Transformer-based pre-trained models (such as BERT, RoBERTa, ALBERT, etc.) perform well in NER tasks. Select one of the pre-trained models suitable for Chinese text NER extraction as the basis for downstream fine-tuning; Model fine-tuning: For the NER task of vehicle text, add a classification layer on the basis of the pre-trained model to identify the entity type in the text (such as vehicle brand, model name, etc.), and use indicators such as accuracy, recall, and F1 score to evaluate model performance.

[0100] Named entity extraction and alias conversion: After obtaining the vehicle model entity extraction model, the trained model is exported to a deployable format (such as ONNX, TensorRT, etc.), and the model is encapsulated as an API service for easy integration into the entire application process. The model is optimized according to the actual application scenario to ensure that it meets the real-time and batch processing requirements. When the newly added verification work text is passed in, the vehicle model word entity is extracted through the model service, and non-standard brands in the text can be extracted for non-work orders. 、Model 、Year , license plate etc. entities; extracted non-standard brands 、Model 、Year , license plate Entities such as , through the alias mapping relationship, map them to standard brands 、Model 、Year , license plate superior, Conversion function for converting nonstandard to standard.

[0101]

[0102]

[0103]

[0104]

[0105] The image and text consistency verification module, after completing the extraction of verification images and verification texts, needs to verify whether the image and text consistency is met from the image dimension and the text dimension, and whether the image and text meet the upload and other content to meet the verification conditions. After completing the training of image and text data, two models are used to perform verification image extraction and verification text extraction on the image data and text data of the verification work order respectively. The image data detection identifies the actual brand, model, year, license plate and other information of the vehicle in the image, which can be recorded as a list The text part extracts the vehicle's brand, model, year, license plate and other information entered in the system text, which can be recorded as a list , used to compare whether the content extracted from the normalized actual uploaded image and the work order text information entered in the work order matches.

[0106]

[0107] At the same time, the threshold is set according to the set proportion of the vehicle body to the picture ,determination Is it satisfied After the conditions, image-text consistency, body ratio threshold, and whether it is a vehicle image label L are met, subsequent verification will continue.

[0108] The product adaptation relationship verification module, after completing the verification of image and text consistency and long-term compliance, needs to verify whether the vehicle model and the product meet the verification logic of the vehicle model for the adapted verification products. On the one hand, it prevents the installation of incompatible products from causing damage to the customer's vehicle, and on the other hand, it prevents merchants from arbitrage by randomly binding products for direct verification, and standardizes verification adaptation. The SKU (stock keeping unit) of the verified product is standard data in the standard library. When the operation and maintenance maintains the data, it will also maintain the vehicle model adaptation data set corresponding to the product. By extracting brand B, model S, and year Y from verification pictures and text data, we determine whether the current model is among the sum of models. We also compare the general attribute information and proprietary attribute information of the verification product with the original model product adaptation relationship library to ensure that the product meets the verification conditions of the current verification model and that there are no errors in the model products.

[0109] The historical image and text duplicate retrieval verification module, in addition to verifying the image and text consistency and product compatibility of the current work order, also needs to verify whether the currently uploaded image has been uploaded repeatedly in the current store and current area within a sliding time interval M. It is used to verify whether there is any repeated historical verification image and text to bypass the verification mechanism and perform false verification.

[0110] In the image and text vector extraction, image vector conversion: you can choose to separate the backbone part of the image recognition model in the vehicle multi-information detection and recognition to extract the feature vector of the image, and perform weighted average and flatten operations on the feature vectors output by different channels to extract the features of the verification work order image. The image is input into the model, and the output is the feature vector E_P of the image; text vector conversion: you can select the output of the last layer of the trained named entity model as the feature vector, convert the text into a feature vector, input the text into the model, and the output can be converted into the feature vector E_T of the text; image and text to Quantity fusion: Since the dimensions of E_P and E_T may be inconsistent, they need to be adjusted to have the same dimension N. The dimension of the feature vector is adjusted by adding a fully connected layer after the two outputs after feature extraction. For example, the dimension of E_P is D_P, and the dimension of E_T is D_T. A fully connected layer F is added to adjust their dimensions to N. After having feature vectors with consistent dimensions, features can be fused by splicing, weighted summation, or feature fusion network. Here, the image and text have the same weight coefficient, and the weight ratio can be set equal. The weight coefficient θ is set to 0.5 for fusion, and finally the image and text fusion feature vector E is obtained.

[0111]

[0112] Image and text vector retrieval engine, vector data storage and vector index construction: You can use databases designed specifically for vector search, such as Elasticsearch, Faiss, Annoy, etc. to store feature vectors. All historical image and text data are extracted through the image and text vector extraction module to extract the corresponding feature vectors and store them in the vector database. An index is then constructed to speed up the search speed of similarity index retrieval. Commonly used similarity measurement methods include cosine similarity and Euclidean distance. You can choose one of them as the similarity measurement calculation method. For two vectors with the same feature length n, and Cosine similarity of It can be expressed as:

[0113]

[0114] After the newly added text and image data of the verification work order is extracted through image and text features, the feature vector E_new of the current work order is obtained, and the image and text retrieval engine is used to perform similarity retrieval and matching on the image and text features within the sliding time period M. The set sliding time period M will be dynamically adjusted based on maintenance, wearing parts, and vehicle mileage, generally ranging from 3 to 12 months. A similarity threshold β is set. When the similarity between the historical feature vector and the feature vector E_new of the current work order is greater than the threshold β, it is determined that there is an act of using repeated historical verification images to bypass the verification mechanism, and it is determined to be a risk of false verification.

[0115] In addition to the verification of the image and text dimensions, the anomaly detection of historical work order risk items also needs to verify the trend changes of the order amount of the corresponding category in the historical orders, based on the actual calibration price. , service fee and floating threshold intervals , formulate abnormal price verification, and the price of verification needs to be within the threshold range ( At the same time, it is necessary to check whether there are abnormal addresses in regional redemption based on the user's historical redemption addresses and different categories, so as to avoid repeated orders in different regions and stores through the same account, and false redemption arbitrage.

[0116] The storage and verification results are output. After completing the verification of the work order's text and image consistency, vehicle model compatibility, historical text and image duplication, and historical work order risk anomaly verification, the multi-dimensional verification of the work order is guaranteed. CCC information at key nodes is stored in the database, with text data stored in a relational database, image data in a cloud storage system, and search vectors in a vector database. This ultimately generates automatic verification results for the work order verification and uses them in downstream applications. Downstream applications can accurately determine the verification condition review results for the vehicle in question based on vehicle model, license plate, compatibility, historical duplication verification, and compliance, and generate approval and rejection reasons, effectively improving the efficiency of the work order verification process.

[0117] The system of the embodiment of the present invention outputs the multi-modal data verification results of the final verification work order image and text, which is used to improve the automated work order review capability, efficiency and accuracy of the background review platform, and can discover potential verification problems in advance, improve the monitoring quality of verification services, and can trace and review work order data to ensure user satisfaction and improve standardization.

[0118] According to the second aspect of the embodiment of the present invention, Figure 4 As shown, a vehicle work order verification device 400 is provided, comprising:

[0119] The result acquisition module 401 is used to obtain the work order image detection result and work order text extraction result corresponding to the vehicle work order to be verified;

[0120] Alias ​​conversion module 402, configured to perform alias conversion processing on the work order text extraction result based on a pre-configured alias library to obtain a work order text extraction result after conversion processing;

[0121] The data verification module 403 is used to perform data consistency verification based on the work order image detection result and the work order text extraction result after conversion processing to obtain the image and text data verification result;

[0122] The review and verification module 404 is used to extract image features and text features from the image and text data verification results, and determine the review and verification results of the vehicle work order to be verified based on the image features and text features.

[0123] Optionally, the vehicle work order verification device 400 further includes a relationship verification module, which is used to:

[0124] The compatibility of the vehicle and the product in the vehicle work order to be verified is verified according to the pre-configured adaptation relationship, and the adaptation relationship verification result of the vehicle work order to be verified is obtained.

[0125] Optionally, the review and verification module 404 is further configured to:

[0126] Perform weighted fusion of image features and text features to obtain an image-text fusion vector;

[0127] Based on the image-text fusion vector and the acquired historical image-text fusion vector, the work order image data and work order text data corresponding to the vehicle work order to be verified are re-verified to obtain a review verification result of the vehicle work order to be verified.

[0128] Optionally, the work order verification device 400 further includes a risk verification module, which is configured to:

[0129] Determine a price verification strategy based on the acquired historical vehicle work order data; verify the amount data in the vehicle work order to be verified according to the price verification strategy to obtain a price risk verification result for the vehicle work order to be verified; wherein the historical vehicle work order data refers to historical data of work orders of the same category or similar categories as the vehicle work order to be verified;

[0130] Alternatively, an address verification strategy is determined based on the acquired historical vehicle work order data; the regional data in the vehicle work order to be verified is verified according to the address verification strategy to obtain an address risk verification result of the vehicle work order to be verified.

[0131] Optionally, the work order verification device 400 further includes an image detection model module, which is used to:

[0132] The work order image detection model is used to detect and process the work order image data corresponding to the vehicle work order to be verified, so as to obtain the work order image detection result corresponding to the vehicle work order to be verified; wherein the work order image detection result includes at least one of the following: vehicle body image detection result, vehicle model image detection result and license plate image detection result; and is also used to: iteratively optimize the work order image detection model according to the work order image detection result.

[0133] Optionally, the work order verification device 400 further includes an entity extraction model module, which is used to:

[0134] The work order entity extraction model is used to perform entity extraction processing on the work order text data corresponding to the vehicle work order to be verified, so as to obtain the work order text extraction result corresponding to the vehicle work order to be verified; wherein the work order text extraction result includes at least one of the following: vehicle model entity extraction result and license plate entity extraction result; and is also used to: iteratively optimize the work order entity extraction model according to the work order text extraction result.

[0135] It should be noted that the specific implementation content of the vehicle work order verification device of the present invention has been described in detail in the above vehicle work order verification method, so the repeated content will not be described again here.

[0136] According to a third aspect of an embodiment of the present invention, an electronic device is provided, comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by one or more processors, the one or more processors implement the method provided by the first aspect of the embodiment of the present invention.

[0137] According to a fourth aspect of an embodiment of the present invention, a computer-readable medium is provided, on which a computer program is stored. When the program is executed by a processor, the method provided by the first aspect of the embodiment of the present invention is implemented.

[0138] According to a fifth aspect of an embodiment of the present invention, a computer program product is provided, including a computer program, which implements the method provided in the first aspect of the embodiment of the present invention when executed by a processor.

[0139] Figure 5 An exemplary system architecture 500 is shown to which the vehicle work order verification method or vehicle work order verification device according to the embodiments of the present application can be applied.

[0140] like Figure 5 As shown, system architecture 500 may include terminal devices 501, 502, 503, a network 504, and a server 505. Network 504 is used to provide a medium for communication links between terminal devices 501, 502, 503 and server 505. Network 504 may include various connection types, such as wired or wireless communication links or fiber optic cables.

[0141] Users can use terminal devices 501, 502, and 503 to interact with server 505 via network 504 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 501, 502, and 503, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only as examples).

[0142] The terminal devices 501 , 502 , and 503 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, and desktop computers.

[0143] Server 505 can be a server that provides various services, such as a backend management server (for example only) that supports shopping websites browsed by users using terminal devices 501, 502, and 503. The backend management server can analyze and process received data such as verification requests for vehicle work orders, and provide feedback (for example, verification results of vehicle work orders—for example only) to the terminal device.

[0144] It should be noted that the vehicle work order verification method provided in the embodiment of the present application is generally executed by the server 505. Accordingly, the vehicle work order verification device is generally set in the server 505.

[0145] It should be understood that Figure 5 The number of terminal devices, networks and servers in the embodiment is merely illustrative. Any number of terminal devices, networks and servers may be provided as required.

[0146] Reference below Figure 6 , which shows a structural diagram of a computer system 600 of a terminal device suitable for implementing an embodiment of the present application. Figure 6 The terminal device shown is merely an example and should not limit the functions and scope of use of the embodiments of the present application.

[0147] like Figure 6 As shown, computer system 600 includes a central processing unit (CPU) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage unit 608 into a random access memory (RAM) 603. Various programs and data required for the operation of computer system 600 are also stored in RAM 603. CPU 601, ROM 602, and RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to bus 604.

[0148] The following components are connected to the I / O interface 605: an input section 606 including a keyboard, mouse, and the like; an output section 607 including displays such as a cathode ray tube (CRT), a liquid crystal display (LCD), and speakers; a storage section 608 including a hard disk; and a communication section 609 including a network interface card such as a LAN card or a modem. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the I / O interface 605 as needed. Removable media 611, such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, is installed in the drive 610 as needed, so that computer programs read from the removable media can be installed in the storage section 608 as needed.

[0149] In particular, according to embodiments disclosed herein, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed herein include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for executing the methods illustrated in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 609 and / or installed from removable media 611. When executed by central processing unit (CPU) 601, the computer program performs the aforementioned functions defined in the system of this application.

[0150] It should be noted that the computer-readable medium described herein may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. Computer-readable storage media may include, for example, but are not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or components, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection having one or more conductors, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, device, or component. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. This propagated data signal may take a variety of forms, including, but not limited to, electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device. Program code embodied on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wireline, optical fiber cable, RF, or any suitable combination thereof.

[0151] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or a part of code, and the above-mentioned module, program segment, or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of the boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0152] The units involved in the embodiments described in the present application can be implemented by software or by hardware. The units described can also be set in a processor. For example, it can be described as: a processor includes a result acquisition module, an alias conversion module, a data verification module and a review verification module. Among them, the names of these modules do not constitute a limitation on the modules themselves under certain circumstances. For example, the result acquisition module can also be described as "a module for obtaining the work order image detection results and work order text extraction results corresponding to the vehicle work order to be verified."

[0153] As another aspect, the present application also provides a computer-readable medium, which may be included in the device described in the above embodiment; or it may exist independently and not be assembled into the device. The above computer-readable medium carries one or more programs. When the above one or more programs are executed by a device, the device includes: obtaining the work order image detection result and the work order text extraction result corresponding to the vehicle work order to be verified; performing alias conversion processing on the work order text extraction result based on a pre-configured alias library to obtain the converted work order text extraction result; performing data consistency verification based on the work order image detection result and the converted work order text extraction result to obtain the image and text data verification result; extracting image features and text features from the image and text data verification result, and determining the review verification result of the vehicle work order to be verified based on the image features and text features.

[0154] A computer program product provided by an embodiment of the present invention includes a computer program. When the computer program is executed by a processor, it implements the vehicle work order verification method in the embodiment of the present invention.

[0155] According to the technical solution of the embodiment of the present invention, it has the following advantages or beneficial effects: by obtaining the work order image detection result and the work order text extraction result corresponding to the vehicle work order to be verified; performing alias conversion processing on the work order text extraction result based on the pre-configured alias library to obtain the work order text extraction result after the conversion processing; performing data consistency verification based on the work order image detection result and the work order text extraction result after the conversion processing to obtain the image and text data verification result; extracting image features and text features from the image and text data verification result, and determining the review verification result of the vehicle work order to be verified based on the image features and text features; thereby, the work order verification can be automatically performed, and the verification cost is reduced, and the verification efficiency and the accuracy of the verification results are improved.

[0156] The above specific embodiments do not constitute a limitation on the scope of protection of this application. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may occur depending on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application shall be included within the scope of protection of this application.

[0157] It should be noted that the acquisition, storage and application of user personal information involved in the technical solution of this disclosure are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

Claims

1. A method for verifying a vehicle work order, characterized in that: include: Obtain the work order image detection results and work order text extraction results corresponding to the vehicle work order to be inspected; Performing alias conversion processing on the work order text extraction result based on a pre-configured alias library to obtain a work order text extraction result after conversion processing; Performing data consistency verification based on the work order image detection result and the work order text extraction result after the conversion process to obtain a picture and text data verification result; Image features and text features are extracted from the image and text data verification results, and the review verification results of the vehicle work order to be verified are determined based on the image features and text features.

2. The method according to claim 1, characterized in that The method further comprises: The compatibility of the vehicle and the commodity in the vehicle work order to be verified is verified according to the pre-configured adaptation relationship, and the adaptation relationship verification result of the vehicle work order to be verified is obtained.

3. The method according to claim 1, characterized in that Determining the review and verification result of the vehicle work order to be verified based on the image features and the text features includes: Performing weighted fusion on the image features and text features to obtain an image-text fusion vector; Based on the image-text fusion vector and the acquired historical image-text fusion vector, the work order image data and the work order text data corresponding to the vehicle work order to be verified are re-verified to obtain a review verification result of the vehicle work order to be verified.

4. The method according to claim 1, characterized in that: The method further comprises: Determine a price verification strategy based on the acquired historical vehicle work order data; verify the amount data in the vehicle work order to be verified according to the price verification strategy to obtain a price risk verification result of the vehicle work order to be verified; wherein the historical vehicle work order data is historical data of work orders of the same category or similar categories of the vehicle work order to be verified; Alternatively, an address verification strategy is determined based on the acquired historical vehicle work order data; the regional data in the vehicle work order to be verified is verified according to the address verification strategy to obtain an address risk verification result of the vehicle work order to be verified.

5. The method according to claim 1, characterized in that Before obtaining the inspection result of the work order image corresponding to the vehicle work order to be inspected, it also includes: Using the work order image detection model, the acquired work order image data corresponding to the vehicle work order to be verified is detected and processed to obtain the work order image detection result corresponding to the vehicle work order to be verified; wherein the work order image detection result includes at least one of the following: a vehicle body image detection result, a vehicle model image detection result, and a license plate image detection result; The method further comprises: The work order image detection model is iteratively optimized according to the work order image detection result.

6. The method according to claim 1, characterized in that Before obtaining the work order text extraction result corresponding to the vehicle work order to be inspected, it also includes: Performing entity extraction processing on the work order text data corresponding to the work order to be verified using the work order entity extraction model to obtain the work order text extraction result corresponding to the work order to be verified; wherein the work order text extraction result includes at least one of the following: vehicle model entity extraction result and license plate entity extraction result; The method further comprises: The work order entity extraction model is iteratively optimized according to the work order text extraction result.

7. A vehicle work order verification device, characterized in that: include: The result acquisition module is used to obtain the work order image detection results and work order text extraction results corresponding to the vehicle work order to be verified; An alias conversion module, used to perform alias conversion processing on the work order text extraction result based on a pre-configured alias library to obtain a work order text extraction result after conversion processing; A data verification module is used to perform data consistency verification based on the work order image detection result and the work order text extraction result after the conversion process to obtain a picture and text data verification result; The review and verification module is used to extract image features and text features from the image and text data verification results, and determine the review and verification results of the vehicle work order to be verified based on the image features and text features.

8. An electronic device, characterized in that: include: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 6.

9. A computer readable medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images