Intelligent distribution method for automobile parts supply chain based on mobile terminal image recognition

Abstract

The application relates to the field of automobile aftermarket information technology, and particularly discloses an intelligent distribution method for an automobile accessory supply chain based on mobile terminal image recognition, which extracts standardized vehicle feature data through mobile terminal multi-mode image acquisition and recognition, analyzes user role types and VIP permissions to determine an accessible data field set, completes accessory query, fuzzy matching, same-platform wildcard expansion, field filtering and multi-level sorting based on a preset vehicle model- accessory association graph, distributes processed accessory information to a mobile terminal and automatically triggers generation of a quotation sheet, initiation of inquiry, group pushing and other supply chain actions according to user roles, simultaneously collects user feedback behaviors, dynamically optimizes graph adaptation and prompts manufacturers to supplement data. The application can improve vehicle identity recognition efficiency and accuracy, realizes differentiated information distribution for multiple roles, adapts the supply chain from accessory query to transaction, forms a data self-optimization mechanism, and can significantly improve the collaborative efficiency of the accessory supply chain and the accuracy of accessory matching.

Description

Technical Field

[0001] This invention relates to the field of automotive aftermarket information technology, and in particular to an intelligent distribution method for the auto parts supply chain based on mobile image recognition. Background Technology

[0002] In the automotive aftermarket parts sourcing and procurement scenario, the industry has long relied on traditional methods such as manual verification, browsing paper catalogs, or telephone inquiries to conduct business. Due to the vast variety of car models on the market, the tens of thousands of parts per vehicle, and the lack of universal compatibility between different models for the same part type, accurately identifying vehicle information is a prerequisite for conducting parts business.

[0003] Traditional methods rely on manual input of VINs and manual extraction of vehicle nameplate information, which is prone to errors and cumbersome information retrieval, directly resulting in low efficiency and accuracy in vehicle identification. Furthermore, existing technologies display parts query results indiscriminately, failing to differentiate information distribution based on the roles, permissions, and business needs of users such as dealers and repair shops. This leads to extremely low information-business matching. In addition, query results can only display basic information and cannot directly trigger subsequent supply chain actions such as quoting, procurement, and social media pushes, causing a disconnect between the query and transaction processes and creating information silos. Moreover, the system lacks a user behavior data feedback mechanism, preventing manufacturers from obtaining query coverage and matching accuracy of parts data, making it difficult to optimize the parts data catalog, and further exacerbating the industry pain points of inaccurate parts matching and incomplete data.

[0004] Therefore, there is an urgent need for an intelligent distribution method for the auto parts supply chain based on mobile image recognition to solve the above problems. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent distribution method for the auto parts supply chain based on mobile terminal image recognition, comprising the following steps: Vehicle images are captured via a mobile device, and standardized vehicle feature data is extracted using multi-modal image recognition. The vehicle feature data includes at least a VIN code or brand and model information. Get the current user's role type and VIP permissions, and determine the set of data fields that the user can access based on the preset permission matrix; Based on a preset vehicle model-parts association graph, a candidate parts list is obtained by querying the vehicle feature data, and the candidate parts list is filtered and sorted according to the data field set. The processed parts information is distributed to mobile devices, and corresponding supply chain actions are triggered based on the user's role. Collect user feedback behavior, optimize the adaptation relationship in the association graph based on the feedback, and prompt the data provider to supplement the data.

[0006] Furthermore, this invention also discloses an intelligent distribution system for the auto parts supply chain based on mobile terminal image recognition, comprising: The extraction module is used to collect vehicle images through a mobile terminal and extract standardized vehicle feature data using multi-modal image recognition. The vehicle feature data includes at least the VIN code or brand and model information. The acquisition module is used to obtain the current user's role type and VIP permissions, and determine the set of data fields that the user can access based on a preset permission matrix. The filtering and sorting module is used to obtain a candidate parts list based on a preset vehicle model-parts association graph and the vehicle feature data, and to filter and sort the candidate parts list according to the data field set. The processing module is used to distribute the processed parts information to the mobile device and trigger corresponding supply chain actions based on the user's role. The optimization module is used to collect user feedback, optimize the adaptation relationships in the association graph based on the feedback, and prompt the data provider to supplement the data.

[0007] Furthermore, the processing module includes: The sending unit is used to display the sorted list of accessories in the form of cards on the mobile device. Each card displays the fields that the current user is allowed to view. For price fields that the user does not have permission to view, a placeholder for inquiry is displayed. The dynamic generation unit is used to dynamically generate corresponding supply chain action buttons below the accessory card based on the user's role; The distribution unit is used to automatically generate a structured inquiry form when a repair shop user triggers a one-click inquiry and push it to the instant messaging group or supply chain collaboration group bound to the repair shop through the API interface. The group's intelligent response node automatically distributes the inquiry request. The quotation generation unit is used to read the inventory quantity and cost price of the selected accessories when the dealer user triggers the generation of a quotation, allows the input of the discount rate, and automatically generates a quotation according to preset rules; The data completion unit is used to trigger the manufacturer's data completion mechanism when the candidate parts list returned by the query on the quotation is empty or the compatibility confidence of all parts is lower than a preset confidence threshold. Record the types of missing parts. When the cumulative number of missing parts reaches a preset threshold, generate a data completion task in the manufacturer's center backend.

[0008] This application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described intelligent distribution method for auto parts supply chain based on mobile terminal image recognition.

[0009] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described intelligent distribution method for auto parts supply chain based on mobile terminal image recognition.

[0010] The beneficial effects of this application are as follows: Firstly, this invention uses mobile multi-mode image acquisition combined with VIN code verification and error correction, and nameplate semantic extraction technology to replace manual data entry, completely solving the problems of easy error and low efficiency in vehicle identification, and greatly improving the convenience and accuracy of identification.

[0011] Secondly, this invention constructs a permission matrix based on user roles and VIP levels, dynamically filters and distributes parts information, achieves accurate matching between dealers and repair shops, eliminates the problem of information redundancy indiscriminate display, and improves business matching efficiency.

[0012] Third, this invention relies on user error correction and query behavior feedback to automatically optimize the vehicle model-parts association graph and trigger manufacturer data completion, forming a data self-optimization closed loop, continuously improving the accuracy of parts matching. At the same time, the system does not require modification of existing ERP and inventory systems, making it lightweight and easy to promote on a large scale in the industry. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of a method flow proposed in an embodiment of this application.

[0014] Figure 2 This is a schematic diagram of the system structure proposed in an embodiment of the present invention.

[0015] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0016] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0017] like Figure 1 As shown, this application provides an intelligent distribution method for auto parts supply chain based on mobile terminal image recognition, including the following steps: S1: Collect vehicle images via mobile device, and extract standardized vehicle feature data using multi-mode image recognition. The vehicle feature data includes at least VIN code or brand and model information. S2: Obtain the current user's role type and VIP permissions, and determine the set of data fields that the user can access based on the preset permission matrix; S3: Based on the preset vehicle model-parts association graph, a candidate parts list is obtained by querying the vehicle feature data, and the candidate parts list is filtered and sorted according to the data field set; S4: Distribute the processed parts information to the mobile terminal and trigger corresponding supply chain actions according to the user role. The supply chain actions include at least one of generating a quotation, initiating an inquiry, and pushing to a group chat robot. S5: Collect user feedback behavior, optimize the adaptation relationship in the association graph based on the feedback, and prompt the data provider to supplement data.

[0018] As described in steps S1-S5 above, in the auto parts supply chain scenario, vehicle identification is a prerequisite for parts matching. The relationship between vehicle models and parts is complex, and the data needs and operational permissions of manufacturers, dealers, and repair shops have clear boundaries. Query results need to be quickly converted into transaction actions, and parts data needs to be continuously corrected based on actual usage feedback. However, if the above links are isolated from each other, problems such as high identification errors, inconsistent information display, inefficient supply chain collaboration, and delayed data updates will occur, which cannot support the efficient operation of auto parts circulation. Therefore, an integrated method is needed to achieve intelligent management and control of the entire chain.

[0019] Traditional auto parts query and distribution technologies rely on manual vehicle information entry, unified data display, offline price inquiries and quotations, and manual maintenance of parts data. These technologies suffer from core flaws such as high vehicle information entry error rates, lack of access control, gaps between query and transaction processes, and untimely data optimization, making it difficult to achieve accurate matching, differentiated distribution, and automated collaboration. This invention integrates mobile image recognition, access control matrix management, graph matching, intelligent distribution, and feedback optimization technologies to form a systematic solution that fundamentally addresses the various problems of traditional technologies.

[0020] This invention extracts standardized vehicle features through mobile image recognition, analyzes user identity and permissions, completes intelligent parts matching based on vehicle model-parts association graph, distributes parts information according to roles and automatically triggers supply chain actions, and collects user feedback to achieve data iteration and optimization. It constructs a fully closed-loop intelligent distribution system for the auto parts supply chain, which includes identification, permissions, matching, distribution and feedback, and achieves the overall goal of accurate vehicle identification, differentiated distribution of parts information, efficient supply chain collaboration and self-optimization of parts data.

[0021] This invention uses mobile image acquisition as the data input entry point, and completes the standardized extraction and verification of vehicle identity through multi-mode image recognition. It constructs a permission matrix based on user roles and VIP levels to determine data access boundaries, and completes accurate matching, wildcard expansion, and permission field filtering based on a pre-built vehicle model-parts association graph. The processed parts information is distributed according to user identity and automatically linked to supply chain business actions such as quoting, inquiry, and group robot push. The entire process records user query and error correction behavior, dynamically corrects the graph adaptation relationship according to thresholds, and simultaneously statistically analyzes data coverage and pushes data to manufacturers to complete the data, forming a fully automated and iterative closed-loop operation mechanism.

[0022] This invention achieves high accuracy in vehicle identification and targeted information distribution through image recognition and access control, establishes a closed loop from query to transaction through graph matching and automated supply chain actions, and enables self-iterative optimization of parts data through a feedback mechanism. Without modifying existing ERP and inventory management systems, it can comprehensively improve the matching efficiency, collaboration capabilities, and data quality of the auto parts supply chain, and significantly reduce human error and supply chain operating costs.

[0023] In one embodiment, step S1, which involves acquiring vehicle images and extracting standardized vehicle feature data via a mobile device, specifically includes: S11: The mobile app offers three image acquisition modes: real-time scanning mode, album import mode, and precise region recognition mode; users can select any mode to obtain the original vehicle image. S12: After preprocessing the acquired image, an OCR recognition algorithm is used to recognize the text information within it. If the recognition result contains 17 characters and conforms to the VIN code encoding rules, the result is used as the VIN code, and the 9th check bit of the VIN code is used for verification: calculate the weighted sum of the first 8 characters, with weighting coefficients of 8, 7, 6, 5, 4, 3, 2, and 1 respectively. Take the remainder of the weighted sum divided by 11. If the remainder is 10, the check bit should be X; otherwise, it is the remainder number. If the actual check bit does not match the calculated result, the VIN code is automatically corrected according to the check bit rules. S13: If the recognition result does not conform to the VIN code rules, it is determined to be a vehicle nameplate image, and fields such as brand, model, displacement, and production year are extracted from it; S14: Match the identified VIN code or nameplate field with the local preset vehicle model database to obtain standardized vehicle feature data, including brand code, model series code, engine model and production year, and output the comprehensive identification confidence level, which is divided into three levels: high, medium and low.

[0024] As described in steps S11-S14 above, by using mobile multi-mode image acquisition, optical character recognition, vehicle identification code verification and error correction, vehicle nameplate field extraction, and local pre-set vehicle model database matching, standardized extraction and accurate identification of vehicle features are achieved, thereby achieving the core goal of improving the convenience and accuracy of vehicle identification.

[0025] The automotive aftermarket features a wide variety of vehicle models, and the compatibility of parts is highly dependent on vehicle identification information. Vehicle feature extraction is a crucial preliminary step in auto parts retrieval. Manually entering vehicle identification numbers is prone to character errors, and manually extracting vehicle nameplate information is cumbersome, directly leading to low efficiency and significant data errors in vehicle identification. This fails to provide reliable data support for accurate subsequent parts matching. Therefore, it is necessary to achieve rapid, standardized, and highly accurate extraction of vehicle features.

[0026] Traditional vehicle identification technologies rely on manual input of vehicle identification codes or manual filling of nameplate information, which suffers from high input error rates, low recognition efficiency, and insufficient data standardization. This invention employs a multi-mode image acquisition combined with optical character recognition and dedicated vehicle identification code verification and error correction, specifically addressing the error and efficiency problems of traditional identification methods.

[0027] The mobile app initially offers three image acquisition methods: real-time scanning, album import, and precise region recognition. Users can choose any mode to acquire the original vehicle image based on their specific needs. This design is suitable for various usage scenarios, including real-time scanning at repair shops, importing historical images from offline albums, and high-precision recognition of selected target areas, enhancing the flexibility and convenience of vehicle image acquisition and providing complete original image data for subsequent recognition processes.

[0028] After preprocessing the acquired raw vehicle images, the system uses an optical character recognition (OCR) algorithm to extract text information from the images. When the extracted text is 17 characters long and conforms to the vehicle identification number (VIN) encoding rules, the text is identified as the VIN and a verification and error correction operation is performed. The specific implementation of verification and error correction involves calculating the weighted sum of the first 8 characters of the VIN, with weighting coefficients set sequentially to 8, 7, 6, 5, 4, 3, 2, and 1. The weighted sum is then divided by 11, and the check digit is X when the remainder is 10; otherwise, the check digit is the number corresponding to the remainder. If the actual check digit does not match the calculated result, the system automatically corrects the VIN according to the check digit rules. This verification algorithm directly corrects character errors generated by optical character recognition, eliminating VIN recognition errors at the source, significantly improving the accuracy of vehicle identification, and preventing subsequent parts matching failures due to VIN errors.

[0029] When the extracted text does not conform to the vehicle identification number (VIN) encoding rules, the system determines that the image is a vehicle nameplate image and extracts four core fields from the image: brand, model, engine displacement, and production year. This can cover recognition scenarios without VIN images, supplementing the dimensions of vehicle feature extraction and ensuring the completeness of vehicle feature data acquisition.

[0030] The system matches the identified vehicle identification number (VIN) or nameplate field with a locally pre-stored vehicle model database to generate standardized vehicle feature data containing brand code, model series code, engine model, and production year. It also outputs high, medium, and low comprehensive recognition confidence levels. The locally pre-stored vehicle model database is a set of standard vehicle model data pre-stored by the system. The matching process converts non-standardized recognition results into standardized data adapted to the auto parts supply chain system. The confidence level classification provides a data reliability reference for subsequent vehicle model-parts association graph queries, ensuring that the output vehicle feature data can be directly used for accurate parts matching. This provides a stable and accurate data foundation for subsequent stages of the intelligent distribution method in the entire auto parts supply chain.

[0031] In one embodiment, step S2, which involves obtaining user roles and VIP permissions and determining the set of accessible data fields, specifically includes: S21: When a user logs in, the mobile app sends an authentication request to the server. The server queries the user's role type from the user center database. The role types include three types: manufacturer, distributor, and repair shop. S22: Further query the type of VIP service package subscribed to by the user, including the basic version, self-provided VIP, and technical service package VIP. Among them, self-provided VIP can view the price and inventory of accessories, while technical service package VIP can view technical parameters and installation videos, but the price field is hidden. S23: Based on the preset permission matrix table, the rows of this table are determined by the combination of role type and VIP type, and the columns correspond to different data fields, including parts brand, OE number, wholesale price, suggested retail price, inventory quantity, technical drawings, installation video, and list of compatible vehicle models; read the set of fields that the current user can access from the permission matrix; S24: For the role of distributor, it is additionally determined whether to allow viewing cross-regional inventory information based on the region code to which the distributor belongs. Cross-regional inventory viewing permission is only granted when the distributor is a provincial general agent.

[0032] As described in steps S21-S24 above, by obtaining the user role type and VIP service package type, and combining them with the preset permission matrix table to determine the data access field set, and by adding cross-regional inventory viewing permission judgment for distributors, differentiated data access control based on user identity is achieved, thus achieving the core goal of accurate information distribution for multiple roles.

[0033] The auto parts supply chain scenario includes three types of users: manufacturers, distributors, and repair shops. The business scenarios and data needs of different users have clear boundaries. Distributors need to view wholesale prices and inventory, repair shops need to view technical parameters and installation videos, and manufacturers need to manage data and view coverage. Undifferentiated display of data will cause information redundancy and leakage of commercial prices, and cannot match the actual business needs of each role. Therefore, it is necessary to strictly limit the scope of accessible data based on user identity and permissions.

[0034] Traditional auto parts information systems employ a uniform data display model, failing to differentiate between user roles and VIP levels. This results in inadequate sensitive data control, poor user experience, and low business matching efficiency. This invention addresses these shortcomings by employing a solution that integrates user identity verification, VIP service package matching, permission matrix lookup, and dealer regional permission tiers.

[0035] When a user logs in, the mobile app sends an authentication request to the server. The server then queries the user's central database to determine the user's role type, which is categorized into three types: manufacturer, distributor, and repair shop. The user central database is a pre-built and maintained user identity information repository that stores user role tags and login credentials. It enables precise identification of the user's basic identity, providing a core basis for subsequent permission resolution.

[0036] The server further queries the user's subscribed VIP service package type, which is divided into three types: Basic, Self-Configured VIP, and Technical Service Package VIP. Self-Configured VIP users can view accessory prices and inventory data, while Technical Service Package VIP users can view technical parameters and installation video data; the price field remains hidden. The above steps refine permissions according to user service levels, achieving precise binding between VIP levels and data permissions.

[0037] The system reads the set of accessible fields based on a preset permission matrix table. The rows of the permission matrix table are determined by a combination of role type and VIP type, while the columns correspond to data fields such as parts brand, OE number, wholesale price, suggested retail price, inventory quantity, technical drawings, installation videos, and a list of compatible vehicle models. This matrix table enables standardized permission configuration and fast table lookup, ensuring accurate and efficient field filtering.

[0038] The system additionally performs regional permission checks on distributor roles, determining whether cross-regional inventory viewing permissions are granted based on the distributor's region code. Only provincial-level general agents can obtain cross-regional inventory viewing permissions. This enables regional control of inventory data, preventing disorderly access to cross-regional inventory information and ensuring the compliant use of supply chain data.

[0039] This step constructs a complete user permission control system through multi-dimensional permission parsing and field filtering, enabling differentiated distribution of information such as dealers viewing prices and inventory and repair shops viewing technical parameters. This improves information matching efficiency, protects commercially sensitive data, and provides the basic data for permission compliance for subsequent parts information display and supply chain action triggering.

[0040] In one embodiment, step S3, which involves querying and filtering the candidate parts list based on the vehicle model-parts association graph, specifically includes: S31: The server pre-builds and maintains a vehicle model-parts association graph. This graph stores four types of nodes: vehicle model, part, OE number, and brand, as well as four types of relationships: adaptable, replaceable, universal, and belong. Each adaptable relationship records the adaptability confidence, with an initial value of 1.0, and records the last update time. S32: Based on the brand and model series in the vehicle feature data obtained in step S1, locate the corresponding model node in the map. If an exact match cannot be found, use edit distance fuzzy matching and take the model node with the highest similarity as the target. If the similarity is less than 0.6, return an empty list. S33: Starting from the target vehicle model node, traverse all accessory nodes connected by adaptation edges to obtain an initial candidate accessory list, and then perform universality expansion: If a part is compatible with other car models, and those other car models belong to the same platform as the target car model, then the part will also be added to the candidate list and marked as "universal recommendation"; S34: Based on the set of accessible fields determined in step S2, filter out fields that are not allowed to be displayed for each candidate accessory, and then sort them according to the following rules: First, sort by reliability of appropriate configuration from high to low. If the reliability of appropriate configuration is the same, sort by historical query popularity from high to low. If the historical query popularity is the same, sort by manufacturer recommendation level from high to low. Finally, take the top 50 parts as the returned results.

[0041] As described in steps S31-S34 above, by pre-constructing a vehicle model-parts association graph and combining it with vehicle feature data, parts query, fuzzy matching fallback, platform compatibility expansion, permission field filtering, and multi-level weighted sorting are completed. This achieves the accurate generation and optimized output of the candidate parts list, thus achieving the core goal of improving the accuracy, comprehensiveness, and query efficiency of parts matching.

[0042] In the automotive parts supply chain, the correspondence between vehicle models and parts presents a complex many-to-many relationship. The same part can be adapted to multiple vehicle models on the same platform. Direct matching through a single vehicle model will miss common parts. At the same time, different user permissions require different data fields to be displayed. Irregular part sorting will significantly reduce the efficiency of user selection. Therefore, it is necessary to rely on structured graphs to achieve complete matching, dynamic filtering and orderly output.

[0043] Traditional parts query systems use two-dimensional data tables for simple matching, lack wildcard expansion logic, fail to filter fields based on user permissions, and have single, fixed sorting rules. This results in incomplete parts matching, redundant information display, and insufficient accuracy of query results. This invention employs a graph database for storage, edit distance fuzzy matching as a fallback, platform-wide wildcard expansion, permission-linked field filtering, and a three-level weighted sorting solution to specifically address the shortcomings of traditional query methods.

[0044] The server pre-builds and maintains a vehicle model-parts relationship graph, stored in a graph database format. This graph includes four types of nodes: vehicle model, part, OE number, and brand, as well as four types of relationships: adaptable, replaceable, wildcard, and belong. Each adaptability relationship records an initial adaptability reliability of 1.0 and its most recent update time. Compared to traditional data tables, graph-based storage is better suited to the complex relationships between vehicle models and parts. The adaptability reliability quantifies the reliability of part-vehicle compatibility, providing a quantitative basis for subsequent sorting and data correction.

[0045] The system locates the corresponding vehicle model node in the map based on the brand and model series in the vehicle feature data. When an exact match cannot be achieved, an edit distance algorithm is used for fuzzy matching, selecting the model node with the highest similarity as the query target. If the similarity is lower than 0.6, an empty list is returned directly. Edit distance fuzzy matching is compatible with minor errors in vehicle feature recognition, and the 0.6 similarity threshold strictly ensures the accuracy of model matching, avoiding subsequent parts queries from failing due to incorrect matching.

[0046] Starting from the target vehicle model node, the system traverses all accessory nodes connected by adaptation edges to generate an initial candidate accessory list. Then, it performs a universality expansion operation: if an accessory is compatible with other vehicle models on the same platform as the target vehicle, the accessory is added to the candidate list and marked as a universally recommended accessory. This platform-wide universality expansion effectively covers common accessories missed by traditional matching methods, expanding the completeness of the candidate list and meeting the accessory selection needs of repair shops and dealers.

[0047] The system filters out fields that users are not authorized to display for each candidate accessory based on the set of data fields accessible to the user. Then, it sorts the accessories according to the following rules: compatibility reliability (highest to lowest), historical query popularity (highest to lowest), and manufacturer recommendation level (highest to lowest). Finally, the top 50 accessories are selected as the returned results. The field filtering and access control system work together, prioritizing the display of highly compatible, frequently used, and officially recommended accessories through a three-level sorting system. The limit of 50 items controls data transmission volume, reduces mobile computing power consumption, and improves page loading and query efficiency.

[0048] By using precise graph matching, universal expansion, permission-linked filtering, and multi-level optimized sorting, we can achieve comprehensive and accurate parts query, adapt to the permission requirements of different user roles, shorten the time spent on parts selection, and provide a high-quality and highly available parts data foundation for subsequent intelligent distribution and action triggering in the supply chain.

[0049] In one embodiment, step S4, which involves distributing component information and triggering supply chain actions, specifically includes: S41: Display the sorted parts list as cards on the mobile device. Each card displays the fields that the current user is allowed to view. For the price field, if the user is a repair shop and has not purchased the self-parts VIP, display the "Request a quote" placeholder. S42: Based on the user's role, dynamically generate different supply chain action buttons below the parts card: for distributors, generate buttons such as "Generate Quotation", "Share with Customers", and "Add to Purchase Vehicle"; for repair shops, generate buttons such as "One-Click Inquiry", "View Nearby Distributors with Stock", and "View Installation Tutorial"; for manufacturers, generate buttons such as "Edit Data" and "View Data Coverage". S43: When a repair shop user clicks "One-click Inquiry", the system automatically extracts the current vehicle characteristics and selected parts information to generate a structured inquiry form. The inquiry form includes the vehicle VIN code, parts OE number, parts brand, required quantity, and expected delivery time. Then, the inquiry form is pushed to the instant messaging group or supply chain collaboration group bound to the repair shop through a preset API interface. The group's intelligent response node automatically distributes the inquiry request. S44: When a dealer clicks "Generate Quotation," the system reads the current inventory and cost price of the selected parts, and allows the dealer to enter the discount rate. The total quoted price is calculated as follows: First, multiply the cost price by 1 and add the dealer's preset markup rate to get the base price. Then, multiply the base price by the required quantity to get the subtotal. Finally, multiply the subtotal by 1 and subtract the discount rate to get the final total price. The markup rate is 30% by default. The generated quotation is in PDF format and includes the dealer's name, contact information, and quotation validity period. S45: If the query returns an empty list of candidate parts, or if the compatibility confidence of all parts is below 0.3, the manufacturer data completion mechanism will be triggered. Record the types of missing parts for this model. When the number of missing parts for the same model reaches 10, a data completion task will be generated in the manufacturer's center backend to prompt the manufacturer to supplement the parts data for this model.

[0050] As described in steps S41-S45 above, by displaying card-style parts information, dynamically generating supply chain action buttons according to user roles, automatically generating and pushing inquiry forms, generating PDF quotation forms according to rules, and triggering the manufacturer data completion mechanism with thresholds, the system achieves accurate distribution of parts information and automated triggering of supply chain business actions, thereby achieving the core objectives of shortening the path to finding transactions, improving supply chain collaboration efficiency, and guiding manufacturers to improve parts data.

[0051] The core need of the auto parts supply chain is to quickly transform parts search results into actual transactions. The cumbersome manual inquiry process of repair shops and the inefficiency of dealers manually creating quotations, coupled with missing parts data that cannot be promptly synchronized with manufacturers, lead to a disconnect between the inquiry and transaction processes, low supply chain collaboration efficiency, and delayed data completion, preventing the formation of a complete business loop. Therefore, it is necessary to deeply integrate information display with business actions to achieve integrated linkage of inquiry, price inquiry, quotation, and data completion.

[0052] Traditional auto parts information systems simply display basic parts information, lacking dedicated operation entry points based on user roles. Quotations rely on manual offline communication, and pricing requires manual document compilation. There's no proactive alert mechanism for missing data, resulting in cumbersome processes, inefficient cross-entity collaboration, and a lack of data closure. This invention employs a role-based interface, automated quotation push notifications, rule-based pricing calculations, and threshold-based data completion triggering to specifically address the process fragmentation and inefficient collaboration inherent in traditional systems.

[0053] The system displays the sorted parts list in card format on the mobile device. Each card only shows data fields that the current user is allowed to access. If the repair shop user has not purchased a self-parts VIP membership, the price field will display a placeholder for inquiry. This display method is strictly linked to the access control system, ensuring both the compliance of information display and simplifying the mobile browsing experience with a card layout, thereby improving the efficiency of user information retrieval.

[0054] The system dynamically generates corresponding supply chain action buttons below the parts card based on the user's role. Dealer users get buttons for generating quotations, sharing with customers, and adding to the purchase vehicle; repair shop users get buttons for one-click inquiry, finding nearby dealers with stock, and viewing installation tutorials; and manufacturer users get buttons for editing data and viewing data coverage. Different roles are matched with dedicated operation buttons, which fits the actual business scenarios of various users. There is no need to manually switch function entry points, reducing operational complexity and improving business triggering efficiency.

[0055] When a repair shop user clicks the one-click inquiry button, the system automatically extracts the current vehicle characteristics and selected parts information, generating a structured inquiry form containing the vehicle's VIN code, part's OE number, part brand, required quantity, and expected delivery time. When the repair shop user triggers the one-click inquiry, the system generates a standardized structured inquiry form based on vehicle characteristic data and parts information. This inquiry form is then pushed to the user's pre-bound instant messaging group or supply chain collaboration group via a preset application programming interface. The intelligent response nodes configured within the group automatically receive and distribute the inquiry information, enabling rapid delivery of inquiry requests. This process replaces the traditional method of manually editing inquiry information and sending messages, achieving standardized generation and automated delivery of inquiry requests, significantly shortening inquiry response time and improving the collaborative efficiency between repair shops and dealers.

[0056] When a dealer clicks the "Generate Quote" button, the system automatically reads the current inventory and cost price of the selected parts. Dealers can also manually input the discount rate. The system ultimately generates a PDF quote containing the dealer's header, contact information, and quote validity period. This calculation method and standardized document generation replace manual price calculations and quotation document preparation, improving dealer quoting efficiency and ensuring a consistent and standardized quotation format.

[0057] When the returned candidate parts list is empty, or when the compatibility reliability of all parts is below 0.3, the system triggers the manufacturer's data completion mechanism. This mechanism records the missing parts types for that vehicle model. When the cumulative number of missing parts for the same model reaches 10, a data completion task is automatically generated in the manufacturer's backend, reminding the manufacturer to supplement the corresponding parts data. The compatibility reliability threshold of 0.3 and the cumulative threshold of 10 missing parts accurately identify data gaps, avoid invalid reminders, and proactively push completion tasks to guide manufacturers in improving data, thereby increasing the success rate of subsequent parts queries from the source.

[0058] This step connects the entire process from parts search to inquiry, quotation, and data completion through role-based information distribution and automated business triggering, enabling efficient collaboration among all entities in the supply chain. At the same time, it establishes a proactive feedback mechanism for data gaps, continuously improves the parts data system, and provides closed-loop operational support for the intelligent distribution method of the entire auto parts supply chain.

[0059] In one embodiment, step S5, which involves collecting user feedback behavior and optimizing the association graph, specifically includes: S51: The server records the following behavioral data for each query operation: query time, vehicle characteristics, returned parts list, parts clicked by the user, whether the user initiated an inquiry or generated a quotation, and whether the user marked "incompatible" or "information error". S52: When a user clicks the "Incompatible" button, the system writes the feedback into the error correction queue, recording the user ID, part ID, and vehicle model ID. The same user's error correction for the same group of vehicle models and parts will only be counted once within 24 hours. S53: The server scans the error correction queue every hour and counts the cumulative number of errors for each vehicle model and part. If the number of errors reaches 3, the adaptation confidence of the adaptation relationship is reduced. Specifically, the original confidence is multiplied by 0.5 to obtain the new confidence. If the new confidence drops below 0.2, the adaptation edge is marked as "pending review" and the adaptation relationship is temporarily hidden from the online query. At the same time, a review task is generated to notify the manufacturer of the part to review it. If the manufacturer confirms that the adaptation is wrong, the adaptation edge is permanently deleted from the graph. If the manufacturer confirms that the adaptation is correct, the confidence is restored to 0.9. S54: Generates a data coverage report weekly, counts the total number of queries and successful returns for each manufacturer's parts, calculates the coverage rate (successful returns divided by total queries), and lists vehicle models and parts combinations with a coverage rate below 50% as "data gaps," which are displayed on the manufacturer center H5 page. Manufacturers supplement the data accordingly. After the manufacturer supplements the data, the system automatically updates the adaptation relationships in the graph and resets the missing count for that vehicle model and parts combination.

[0060] As described in steps S51-S54 above, by recording all user query behavior, standardizing error correction counting rules, dynamically adjusting the appropriate configuration reliability according to thresholds, and regularly generating data coverage reports, the system achieves automated optimization of the vehicle model-parts association graph and accurate prompts for manufacturer data gaps, thereby achieving the core objectives of data self-optimization, improving parts matching accuracy, and perfecting the supply chain data ecosystem.

[0061] The vehicle model-parts association map in the auto parts supply chain needs continuous iteration and updates. Error corrections and query gaps identified during actual user use provide the most accurate data for optimization. Manual review and error correction are inefficient, data coverage cannot be quantified, and manufacturers cannot accurately identify their own data shortcomings. This leads to distorted map adaptation relationships and persistent data gaps, directly reducing the reliability of parts queries and matching. Therefore, an automated mechanism for collecting user feedback, map correction, and data gap feedback is needed.

[0062] Traditional auto parts data systems passively store query records, lack standardized processing procedures for user error correction feedback, cannot automatically correct adaptation relationships, and do not establish data coverage statistics and manufacturer push mechanisms. This results in problems such as delayed data updates, reliance on manual error correction, and inaccurate identification of data gaps. This invention employs a technical solution that specifically addresses the lag and inefficiency of traditional data maintenance by implementing full behavior recording, error correction frequency threshold control, dynamic confidence decay, and periodic quantitative report generation.

[0063] The server records complete behavioral data for each query operation, including query time, vehicle characteristics, returned parts list, parts clicked by the user, whether the user initiated an inquiry or generated a quote, and whether the user marked the item as incompatible or with incorrect information. This comprehensive behavioral record provides raw data support for subsequent error correction statistics, graph optimization, and data report generation, ensuring that feedback processing and data statistics have a reliable data source.

[0064] When a user clicks the "Incompatible" button, the system writes the feedback to the error correction queue, simultaneously recording the user ID, part ID, and vehicle model ID. Error corrections from the same user for the same vehicle model and part are counted only once within 24 hours. This rule excludes invalid error correction data submitted repeatedly by the same user, ensuring the accuracy of error correction statistics and preventing incorrect corrections of compatibility relationships due to repeated operations.

[0065] The server scans the error correction queue hourly, accumulating the number of corrections for each vehicle model and part. When the number of corrections reaches 3, the original confidence level of the adaptation relationship is multiplied by 0.5 to obtain a new confidence level. When the new confidence level drops below 0.2, the system marks the adaptation edge as pending review and temporarily hides the adaptation relationship from the online query. Simultaneously, a review task is generated to notify the manufacturer of the part for verification. If the manufacturer confirms the adaptation error, the adaptation edge is permanently deleted from the graph; if the manufacturer confirms the adaptation is correct, the confidence level is restored to 0.9. The 3-correction threshold, the 0.5 confidence level decay coefficient, and the 0.2 hiding threshold accurately identify high-frequency erroneous adaptation relationships. The temporary hiding mechanism prevents erroneous data from affecting user queries, the manufacturer verification process ensures the accuracy of graph correction, and dynamic confidence level adjustment achieves quantitative optimization of adaptation relationships.

[0066] The system generates a weekly data coverage report, calculating the total number of queries and successful returns for each manufacturer's parts. Coverage is calculated by dividing the number of successful returns by the total number of queries. The system identifies vehicle models and parts combinations with coverage below 50% as data gaps, displaying this information to manufacturers via a manufacturer center H5 page. After manufacturers supplement the data, the system automatically updates the matching relationships in the data map and resets the missing count for that vehicle model and parts combination. The formula for calculating manufacturer parts data coverage is as follows: ; Among them, the This indicates the coverage rate of manufacturer's parts data. This indicates the number of times the manufacturer's parts query has successfully returned results. This indicates the total number of times a manufacturer's parts are queried, with a weekly statistical cycle and a 50% coverage threshold. It allows for regular quantification of data quality, accurate identification of manufacturers' data shortcomings, and an automatic update and reset mechanism that forms a closed loop for data completion, continuously improving the completeness and matching success rate of the graph data.

[0067] This step establishes a self-circulating system for data collection, error correction, optimization, and completion through automated processing of user feedback, dynamic correction of correlation maps, quantitative statistics of data quality, and manufacturer push notifications. This continuously improves the accuracy and completeness of vehicle model and parts correlation maps, solves the problems of parts matching errors and data gaps from the source, provides stable and reliable data assurance for the intelligent distribution method of the entire auto parts supply chain, and promotes the healthy iteration of the supply chain data ecosystem.

[0068] like Figure 2 As shown, this invention also discloses an intelligent distribution system for the auto parts supply chain based on mobile terminal image recognition, comprising: The extraction module is used to collect vehicle images through a mobile terminal and extract standardized vehicle feature data using multi-modal image recognition. The vehicle feature data includes at least the VIN code or brand and model information. The acquisition module is used to obtain the current user's role type and VIP permissions, and determine the set of data fields that the user can access based on a preset permission matrix. The filtering and sorting module is used to obtain a candidate parts list based on a preset vehicle model-parts association graph and the vehicle feature data, and to filter and sort the candidate parts list according to the data field set. The processing module is used to distribute the processed parts information to the mobile device and trigger corresponding supply chain actions based on the user's role. The optimization module is used to collect user feedback, optimize the adaptation relationships in the association graph based on the feedback, and prompt the data provider to supplement the data.

[0069] In one embodiment, the processing module includes: The sending unit is used to display the sorted list of accessories in the form of cards on the mobile device. Each card displays the fields that the current user is allowed to view. For price fields that the user does not have permission to view, a placeholder for inquiry is displayed. The dynamic generation unit is used to dynamically generate corresponding supply chain action buttons below the accessory card based on the user's role; The distribution unit is used to automatically generate a structured inquiry form when a repair shop user triggers a one-click inquiry and push it to the instant messaging group or supply chain collaboration group bound to the repair shop through the API interface. The group's intelligent response node automatically distributes the inquiry request. The quotation generation unit is used to read the inventory quantity and cost price of the selected accessories when the dealer user triggers the generation of a quotation, allows the input of the discount rate, and automatically generates a quotation according to preset rules; The data completion unit is used to trigger the manufacturer's data completion mechanism when the candidate parts list returned by the query on the quotation is empty or the compatibility confidence of all parts is lower than a preset confidence threshold. Record the types of missing parts. When the cumulative number of missing parts reaches a preset threshold, generate a data completion task in the manufacturer's center backend.

[0070] This application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described intelligent distribution method for auto parts supply chain based on mobile terminal image recognition.

[0071] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described intelligent distribution method for auto parts supply chain based on mobile terminal image recognition.

[0072] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in this application and in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0073] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0074] The above description is merely a preferred embodiment of the present invention and does not limit the scope of this application. Any equivalent results or equivalent process transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the scope of protection of this application.

Claims

1. A method for intelligent distribution of auto parts in the supply chain based on mobile image recognition, characterized in that, Includes the following steps: Vehicle images are captured via a mobile device, and standardized vehicle feature data is extracted using multi-modal image recognition. The vehicle feature data includes at least a VIN code or brand and model information. Get the current user's role type and VIP permissions, and determine the set of data fields that the user can access based on the preset permission matrix; Based on a preset vehicle model-parts association graph, a candidate parts list is obtained by querying the vehicle feature data, and the candidate parts list is filtered and sorted according to the data field set. The processed parts information is distributed to mobile devices, and corresponding supply chain actions are triggered based on the user's role. Collect user feedback behavior, optimize the adaptation relationship in the association graph based on the feedback, and prompt the data provider to supplement the data.

2. The intelligent distribution method for auto parts supply chain based on mobile terminal image recognition according to claim 1, characterized in that, The extraction of standardized vehicle feature data using multi-modal image recognition specifically includes: The mobile app offers multiple image acquisition modes, allowing users to select any mode to obtain the original vehicle image; After preprocessing the acquired images, an OCR recognition algorithm is used to identify the text information within them, and verification and correction are performed according to the VIN code encoding rules. When the recognition result does not conform to the VIN code rules, it is determined to be a vehicle nameplate image, from which the brand, model, displacement, and production year fields are extracted; The identified VIN code or nameplate field is matched with a pre-set vehicle model database to obtain standardized vehicle feature data, and a comprehensive recognition confidence score is output.

3. The intelligent distribution method for auto parts supply chain based on mobile terminal image recognition according to claim 1, characterized in that, The step of determining the set of data fields that the user can access based on a preset permission matrix specifically includes: When a user logs in, the mobile app sends an authentication request to the server. The server then queries the user's role type from the user center database. The role type includes manufacturer, distributor, and repair shop. To find out the type of VIP service package the user has subscribed to, different VIP types correspond to different data access permissions; Based on the preset permission matrix table, read the set of accessible data fields corresponding to the current user role type and VIP type combination from the table; For distributors, their permission to view cross-regional inventory information is determined additionally based on their region code.

4. The intelligent distribution method for auto parts supply chain based on mobile terminal image recognition according to claim 1, characterized in that, The step of filtering and sorting the candidate parts list based on the data field set specifically includes: The server-side pre-builds and maintains a vehicle model-parts association graph; Based on the brand and model series in the vehicle feature data, locate the corresponding model node in the model-parts association graph. If a precise match cannot be made, fuzzy matching is used. When the similarity is lower than the preset threshold, an empty list is returned. Starting from the target vehicle model node, traverse all accessory nodes connected by adaptation edges to obtain an initial candidate accessory list, and perform universality expansion to add accessories that are compatible with other vehicle models on the same platform to the candidate list; Each candidate accessory is filtered by field based on the set of accessible data fields, and then sorted according to a preset sorting rule before being returned.

5. The intelligent distribution method for auto parts supply chain based on mobile terminal image recognition according to claim 1, characterized in that, The triggering of corresponding supply chain actions based on user roles specifically includes: The sorted list of accessories is displayed as cards on the mobile device. Each card shows the fields that the current user is allowed to view. For price fields that the user does not have permission to view, a placeholder for "need to inquire" is displayed. Based on the user's role, dynamically generate corresponding supply chain action buttons below the accessory card; When a repair shop user triggers a one-click inquiry, a structured inquiry form is automatically generated and pushed to the instant messaging group or supply chain collaboration group bound to the repair shop via API interface. The group's intelligent response node then automatically distributes the inquiry request. When a dealer user triggers the generation of a quotation, the system reads the inventory quantity and cost price of the selected accessories, allows the input of discount rates, and automatically generates a quotation according to preset rules. When the query returns an empty list of candidate parts or the compatibility confidence of all parts is lower than the preset confidence threshold, the manufacturer data completion mechanism is triggered: Record the types of missing parts. When the cumulative number of missing parts reaches a preset threshold, generate a data completion task in the manufacturer's center backend.

6. The intelligent distribution method for auto parts supply chain based on mobile terminal image recognition according to claim 1, characterized in that, The step of optimizing the adaptation relationships in the association graph based on feedback and prompting the data provider to supplement data specifically includes: The server records behavioral data for each query operation. This behavioral data includes query time, vehicle characteristics, returned parts list, parts clicked by the user, whether the user initiated supply chain actions, and whether the user marked the parts as incompatible or incorrect. When a user marks an incompatibility, the feedback is written to the error correction queue, recording the user identifier, part identifier, and vehicle model identifier, and duplicate error corrections from the same user are deduplicated. Regularly scan the error correction queue, and count the cumulative number of error corrections for each group of vehicle models and parts. When the number of error corrections reaches the preset error correction threshold, reduce the compatibility reliability of the corresponding compatibility relationship. When the confidence level drops below the preset hiding threshold, the adaptation edge is marked as pending review and temporarily hidden, and the manufacturer is notified for review. Regularly generate data coverage reports, count the total number of queries and successful returns for each manufacturer's parts, calculate the coverage rate, and list the vehicle-part combinations with coverage rates below the preset coverage rate threshold as data gaps and display them through the manufacturer center. After the manufacturer supplements the data, the map is automatically updated and the missing count is reset.

7. A smart distribution system for auto parts supply chain based on mobile terminal image recognition, characterized in that, include: The extraction module is used to collect vehicle images through a mobile terminal and extract standardized vehicle feature data using multi-modal image recognition. The vehicle feature data includes at least the VIN code or brand and model information. The acquisition module is used to obtain the current user's role type and VIP permissions, and determine the set of data fields that the user can access based on a preset permission matrix. The filtering and sorting module is used to obtain a candidate parts list based on a preset vehicle model-parts association graph and the vehicle feature data, and to filter and sort the candidate parts list according to the data field set. The processing module is used to distribute the processed parts information to the mobile device and trigger corresponding supply chain actions based on the user's role. The optimization module is used to collect user feedback, optimize the adaptation relationships in the association graph based on the feedback, and prompt the data provider to supplement the data.

8. The intelligent auto parts supply chain distribution system based on mobile terminal image recognition according to claim 7, characterized in that, The processing module includes: The sending unit is used to display the sorted list of accessories in the form of cards on the mobile device. Each card displays the fields that the current user is allowed to view. For price fields that the user does not have permission to view, a placeholder for inquiry is displayed. The dynamic generation unit is used to dynamically generate corresponding supply chain action buttons below the accessory card based on the user's role; The distribution unit is used to automatically generate a structured inquiry form when a repair shop user triggers a one-click inquiry and push it to the instant messaging group or supply chain collaboration group bound to the repair shop through the API interface. The group's intelligent response node automatically distributes the inquiry request. The quotation generation unit is used to read the inventory quantity and cost price of the selected accessories when the dealer user triggers the generation of a quotation, allows the input of the discount rate, and automatically generates a quotation according to preset rules; The data completion unit is used to trigger the manufacturer's data completion mechanism when the candidate parts list returned by the query on the quotation is empty or the compatibility confidence of all parts is lower than a preset confidence threshold. Record the types of missing parts. When the cumulative number of missing parts reaches a preset threshold, generate a data completion task in the manufacturer's center backend.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images