An accessory identification method and apparatus

Abstract

This application discloses a method for parts identification. The method first acquires image information of the part to be identified and extracts corresponding part information. Then, it uses this information to query a preset electronic parts catalog. Based on the query results, it flexibly uses an object storage service image library to match candidate part images. Finally, it determines the target part based on the degree of matching between the candidate part information and the part information to be identified. This method integrates the advantages of accurate catalog querying and image visual matching, overcoming the shortcomings of single identification schemes. It comprehensively mines key part information to reduce misjudgments, uses layered processing to balance recognition efficiency and adaptability to complex scenarios, reduces the mismatch rate of similar parts, and quantifies matching judgments to improve the accuracy of results. This achieves efficient and accurate parts identification, meeting the high reliability and efficiency requirements of the automotive aftermarket, reducing user time costs, and improving business transaction efficiency and data update timeliness.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a method and apparatus for identifying accessories. Background Technology

[0002] The automotive aftermarket, as a crucial component of the automotive industry chain, encompasses multiple business segments including vehicle repair, maintenance, parts supply, transactions, and parts parameter queries. It involves a wide range of data sources, including vehicle manufacturer parts catalogs, dealer inventory lists, repair work orders, data from parts e-commerce platforms, vehicle photos, and technical documents. However, this data is generally unstructured, heterogeneous, and inconsistently formatted. Significant differences in description methods, encoding rules, and language systems across different data sources lead to substantial challenges in data processing. Currently, the industry mainly relies on manual or traditional identification methods to handle parts identification needs. Manual processing is inefficient, and data cleaning rules are difficult to standardize. Users need to spend a lot of time locating specific parts information when consulting documents or repair manuals. Among traditional identification methods, optical character recognition solutions have limited applicability and insufficient recognition accuracy when dealing with scenarios such as no inscriptions, worn inscriptions, occlusions, and complex tabular data. Image matching solutions based on general embedding models have a high mismatch rate when dealing with parts that look similar but have different functions. Visual-language joint recognition solutions suffer from insufficient adaptability, resulting in recognition accuracy that fails to meet industry requirements. These problems not only easily lead to misjudgments in parts identification but also affect transaction efficiency and the timeliness of data updates, making it difficult to meet the actual needs of the automotive aftermarket for highly reliable and efficient parts identification. Summary of the Invention

[0003] This invention provides a method and apparatus for parts identification, which can achieve high efficiency and accuracy in parts identification, meet the automotive aftermarket's demand for high reliability and efficiency in parts identification, reduce the time cost for users to locate parts information, and improve the transaction efficiency and timeliness of data updates in related businesses.

[0004] In a first aspect, the present invention provides a method for identifying accessories, the method comprising: Obtain image information of the accessory to be identified; Information is extracted from the image information of the accessory to be identified to obtain the accessory information corresponding to the accessory to be identified; Using the accessory information corresponding to the accessory to be identified, a preset electronic catalog of accessories is queried to obtain the query results; If the number of candidate accessories matching the accessory information corresponding to the accessory to be identified is 0, equal to or greater than 2, then a candidate accessory image matching the image information of the accessory to be identified is determined from the preset object storage service image library. Based on the candidate accessory image, determine the accessory information of the candidate accessory corresponding to the candidate accessory image; If the matching degree between the accessory information of the candidate accessory corresponding to the candidate accessory image and the accessory information corresponding to the accessory to be identified is greater than a preset threshold, then the candidate accessory corresponding to the candidate accessory image is taken as the target accessory corresponding to the image information of the accessory to be identified.

[0005] Secondly, the present invention provides an accessory identification device, the device comprising: The information acquisition unit is used to acquire image information of the accessory to be identified; An information extraction unit is used to extract information from the image information of the accessory to be identified, and obtain accessory information corresponding to the accessory to be identified; The result query unit is used to query a preset electronic catalog of accessories using the accessory information corresponding to the accessory to be identified, and obtain the query result. An image determination unit is used to determine, if the number of candidate accessories matching the accessory information corresponding to the accessory to be identified is 0, equal to or greater than 2, from a preset object storage service image library; An information determination unit is used to determine the accessory information of the candidate accessory corresponding to the candidate accessory image based on the candidate accessory image; The accessory determination unit is configured to, if the degree of matching between the accessory information of the candidate accessory corresponding to the candidate accessory image and the accessory information corresponding to the accessory to be identified is greater than a preset threshold, then use the candidate accessory corresponding to the candidate accessory image as the target accessory corresponding to the image information of the accessory to be identified.

[0006] Thirdly, the present invention provides a readable medium including executable instructions, which, when executed by a processor of an electronic device, cause the electronic device to perform any of the methods described in the first aspect.

[0007] Fourthly, the present invention provides an electronic device including a processor and a memory storing execution instructions, wherein when the processor executes the execution instructions stored in the memory, the processor performs the method as described in any of the first aspects.

[0008] As can be seen from the above technical solution, the method provided by the present invention first acquires the image information of the accessory to be identified, then extracts information from the image information to obtain the corresponding accessory information, and then uses the accessory information to query a preset electronic catalog of accessories. Based on the query results, it flexibly chooses whether to determine candidate accessory images from a preset object storage service image library, and then determines the corresponding accessory information based on the candidate accessory images and determines the target accessory by judging the degree of matching. This derivation process enables the method to fully combine the advantages of accurate querying of the electronic catalog of accessories with the visual matching capabilities of the object storage service image library, effectively making up for the shortcomings of existing single recognition schemes; wherein, by extracting information from the image information This system can comprehensively uncover key information related to parts, providing sufficient basis for subsequent matching and reducing misjudgments caused by missing information. By first querying the electronic parts catalog and then selectively using the image library for matching, it ensures recognition efficiency when there is clear matching information, and solves the recognition problem in scenarios with no matching or multiple matching, reducing the mismatch rate of similar-looking parts. Through quantitative judgment of the degree of matching, the accuracy of the recognition results is further improved, ultimately achieving high efficiency and precision in parts identification. This meets the automotive aftermarket's demand for high reliability and efficiency in parts identification, while reducing the time cost for users to locate parts information and improving the transaction efficiency and timeliness of data updates in related businesses.

[0009] The further effects of the aforementioned non-conventional preferred method will be explained below in conjunction with specific embodiments. Attached Figure Description

[0010] To more clearly illustrate the embodiments of the present invention or the existing technical solutions, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a flowchart illustrating an accessory identification method according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating an accessory identification method according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of an accessory identification device according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0013] Various non-limiting embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0014] See Figure 1 This illustration shows a component identification method according to an embodiment of the present invention. In this embodiment, the method may include, for example, the following steps: S101: Obtain image information of the accessory to be identified.

[0015] The purpose of this step is to collect image data that can be used for accessory identification, providing a foundation for subsequent information extraction and matching. In one implementation, the system first receives accessory-related data uploaded by the user through an interactive interface. This accessory-related data includes image files or document files containing accessory images. Then, the data is parsed to extract accessory image information. This accessory image information refers to image data that reflects the appearance, structure, and markings of the accessory, including images taken from different angles and exploded images of the accessory after disassembly (exploded images of disassembled accessories are also called exploded views of parts; both are the same concept, referring to images showing the relative positions of the components after disassembly). Finally, this accessory image information can be used as the image information of the accessory to be identified.

[0016] For example, the specific process is as follows: First, the system receives data related to the parts to be identified uploaded by the user through the interactive interface. This data includes image files (such as real photos of the parts in JPG or PNG format) or document files containing part images (such as PDF format repair manuals or parts catalogs). This data upload method is similar to... Figure 2This corresponds to the "PDF document or image input" step following "Customer Visit". Subsequently, the system activates the document format parsing module to analyze the format of the user-uploaded data related to the parts to be identified. For independent image files, the image data is directly extracted; for document files containing part images, the part image data is separated using a document parsing algorithm, ultimately extracting the part image information. This part image information includes images of the part from different angles (e.g., front, side, top views) and exploded images of the disassembled part, ensuring a comprehensive reflection of the part's appearance and structural features. Finally, the parsed and extracted part image information is used as the image information for the parts to be identified in subsequent processing, proceeding to the next step.

[0017] S102: Extract information from the image information of the accessory to be identified to obtain accessory information corresponding to the accessory to be identified.

[0018] Parts information: refers to the set of data used to identify and describe the attributes of parts, including parts identification (such as part number), specifications, parts description text information, category attributes, component names, etc.

[0019] This step extracts and integrates text and structured data from the image to uncover key attribute information of the component to be identified, providing a basis for subsequent query matching. Figure 2 The relevant processing steps include "text extraction (OCR)," "table data extraction (large model intelligent recognition)," and "intelligent cutting of part drawings."

[0020] As an example, an image region segmentation algorithm can be used to segment the image information of the accessory to be identified into blocks, locating and extracting the target image region containing accessory identification text. Specifically, an image region segmentation algorithm (such as a threshold-based segmentation or edge detection-based segmentation algorithm) can be used to segment the image information of the accessory to be identified into blocks. This algorithm can divide the image into multiple independent regions based on differences in pixel grayscale, texture, and other features, thereby locating and extracting the target image region containing accessory identification text (such as inscriptions, labels, or printed text), avoiding interference from non-text regions in subsequent recognition.

[0021] Then, an optical character recognition (OCR) module can be used to recognize and collect the text information in the target image area to extract the accessory attribute information. This accessory attribute information includes accessory identification, specifications, and accessory description text. Specifically, the OCR module can be activated to recognize and collect the text information in the target image area to extract the accessory attribute information. This accessory attribute information includes accessory identification (such as part number, also known as PartNumber; both are the same concept, referring to a unique code used to identify the accessory), specifications (such as dimensions, material, and rated parameters), and accessory description text (such as accessory function description and compatible vehicle models).

[0022] Next, structured data detection can be performed on the image information of the accessory to be identified. If structured data in tabular form is detected, a preset table recognition model is called to parse the field content in the table and convert the parsed field data into structured information in a unified format. Specifically, structured data detection can be performed on the image information of the accessory to be identified to determine whether there is structured data in tabular form (such as...) in the image. Figure 2 The "parts catalog data" table mentioned in the document). If structured data in tabular form is detected, a preset table recognition model is invoked (i.e., Figure 2 The model corresponding to "Table Data Extraction (Large Model Intelligent Recognition)" can automatically identify the row and column boundaries, headers and content of a table, parse the field content in the table, and convert the parsed field data into structured information in a unified format (such as JSON format) to ensure the consistency of data format.

[0023] Finally, the extracted part attribute information can be associated and bound with preset standard fields to form the part information corresponding to the part to be identified. Specifically, the part attribute information extracted in the above steps can be associated and bound with preset standard fields (such as "PartNumber", "PartName", "Specifications", "Compatible Vehicle Models", etc.). For example, the identified part number can be bound with the "PartNumber" field, and the specifications can be bound with the corresponding standard fields. Finally, the part information corresponding to the part to be identified can be formed, and this part information will serve as the core basis for subsequent query matching.

[0024] S103: Using the accessory information corresponding to the accessory to be identified, query the preset electronic catalog of accessories to obtain the query results.

[0025] An electronic parts catalog refers to a standardized collection of multi-source structured parts data stored locally or in the cloud for quick querying and matching of parts information.

[0026] This step quickly filters and matches candidate parts by comparing the parts information with a preset catalog. Figure 2 The "standard image retrieval" and "electronic catalog matching" steps aim to improve recognition efficiency. In one implementation, the generation of the preset electronic catalog of parts may include the following steps: collecting multi-source structured data of parts; standardizing the description specifications, encoding systems, and language formats of data from different sources to obtain standardized data; and storing the standardized data in a local storage medium or a cloud storage system as the preset electronic catalog of parts.

[0027] In other words, in one implementation method, it is first necessary to pre-build the preset electronic parts catalog. The specific generation method is as follows: First, collect multi-source structured parts data. The data sources include vehicle manufacturer parts catalogs, dealer inventory lists, repair work orders, structured data from parts e-commerce platforms, etc. Figure 2 The first step corresponds to the data source in "Data Entry (Data Vectorization)". The second step involves standardizing the collected multi-source data due to inconsistencies in description standards, coding systems, and language formats across different sources. This includes converting part numbers with different coding rules into industry-standard coding formats, translating component names from different languages ​​into a unified language, and standardizing the units and expressions of specifications from different formats. The result is standardized data. The third step stores this standardized data on local storage media or a cloud storage system, creating a pre-defined electronic catalog of components. This catalog supports rapid retrieval and comparison and is continuously updated through a subsequent data incremental expansion mechanism.

[0028] After the electronic parts catalog is built, the specific process for querying and matching is as follows: The system extracts the core query fields (such as parts identification / part number) from the parts information corresponding to the parts to be identified, and uses these fields as search keywords to perform a precise query on the preset electronic parts catalog. If there is a record in the electronic catalog that completely matches the keyword, the corresponding candidate parts information is returned; if there is no matching record or there are multiple matching records (i.e., the number of candidate parts is equal to or greater than 2), the corresponding query results are returned, and the process proceeds to the next step.

[0029] S104: If the number of candidate accessories matching the accessory information corresponding to the accessory to be identified is 0, equal to or greater than 2, then a candidate accessory image matching the image information of the accessory to be identified is determined from the preset object storage service image library.

[0030] Object Storage Service Image Library (OSS Image Library for short) is an object storage service system used to store massive amounts of component image samples and related attribute information, providing data support for image matching.

[0031] When the electronic catalog search for accessories cannot yield a unique match, i.e., if the number of candidate accessories matching the accessory information to be identified is 0, equal to, or greater than 2, this step filters candidate accessory images through image feature matching. Figure 2 The relevant steps of "image embedding", "vector analysis" and "OSS image library" retrieval are used to solve the recognition problems in complex scenarios such as no inscriptions and multiple matches.

[0032] In one implementation, a search request can be sent to a pre-defined object storage service image library to retrieve a set of accessory image samples stored in the image library. Specifically, a search request can be sent to a pre-defined object storage service image library (OSS image library) to retrieve a set of accessory image samples stored in that image library. The object storage service image library stores a massive number of standardized accessory image samples, each sample being associated with corresponding accessory attribute information, providing ample data support for image matching.

[0033] Then, the image information of the accessory to be identified can be preprocessed with feature enhancement to obtain the preprocessed image information of the accessory to be identified. Feature enhancement preprocessing refers to the process of removing interference and extracting features from the image. Specifically, it involves removing useless information such as background interference and noise from the image, while enhancing the recognizability of key features such as the accessory's outline and markings, thereby improving the accuracy of subsequent feature extraction. The feature enhancement preprocessing is a processing method that removes interference information from the image and highlights the key features of the accessory. Specifically, the image information of the accessory to be identified can be preprocessed with feature enhancement to obtain the preprocessed image information of the accessory to be identified. The feature enhancement preprocessing specifically involves: removing noise interference from the image through image denoising algorithms (such as Gaussian filtering and median filtering), optimizing image quality through brightness equalization and contrast enhancement algorithms, and highlighting key features such as the accessory's outline and texture through edge enhancement algorithms to ensure the accuracy of subsequent feature extraction.

[0034] Next, a convolutional neural network encoder can be used to extract features from the preprocessed image information of the accessory to be identified, obtaining a high-dimensional feature embedding vector corresponding to the image information of the preprocessed accessory. The high-dimensional feature embedding vector refers to the high-dimensional data vector generated after feature extraction of the image by the convolutional neural network encoder, which can represent the key features of the image. Specifically, a convolutional neural network encoder (such as the EfficientNet model or other dedicated convolutional encoder models) can be used to extract features from the preprocessed image information of the accessory to be identified. This encoder can automatically learn high-order features in the image, converting the image data into a high-dimensional feature embedding vector that can represent its core features.

[0035] Next, the cosine similarity between the high-dimensional feature embedding vector and the feature vectors corresponding to each sample image in the accessory image sample set can be calculated. Specifically, the cosine similarity between the high-dimensional feature embedding vector and the feature vectors corresponding to each sample image in the accessory image sample set can be calculated. Cosine similarity measures the angle between two vectors; the smaller the angle, the higher the similarity, effectively reflecting the degree of feature similarity between the accessory image to be identified and the sample images.

[0036] Finally, the cosine similarity scores of each sample image can be sorted from high to low, and the top-ranked sample images can be selected as candidate accessory images that match the image information of the accessory to be identified. Specifically, the cosine similarity scores of each sample image can be sorted from high to low, and the top-ranked sample images (such as the top 1-20, which can be adjusted according to actual accuracy requirements) can be selected as candidate accessory images that match the image information of the accessory to be identified, forming a candidate image set, and proceeding to the next step.

[0037] S105: Based on the candidate accessory image, determine the accessory information of the candidate accessory corresponding to the candidate accessory image.

[0038] This step analyzes the images of candidate parts to extract corresponding part attribute information, providing data support for subsequent matching degree judgment. Figure 2 This refers to a portion of the "VLM vectorized retrieval (part number matching, image analysis)" process.

[0039] In one implementation, image recognition and semantic analysis can be performed on the candidate part image to obtain the part information of the candidate part corresponding to the candidate part image. The part information of the candidate part corresponding to the candidate part image includes the category attribute, component name, and key parameter information of the candidate part. Specifically, the system can call the image recognition and semantic analysis module to process the candidate part image—first, extracting structural features and identification features from the candidate part image using an image recognition algorithm; then, interpreting the image features using a semantic analysis model to finally obtain the part information of the candidate part corresponding to the candidate part image. The part information of the candidate part includes the category attribute (e.g., engine part, chassis part), component name (e.g., water pump connector, bearing), and key parameter information (e.g., size, material, compatible model), ensuring that this information is comparable to the part information corresponding to the part to be identified.

[0040] S106: If the matching degree between the accessory information of the candidate accessory corresponding to the candidate accessory image and the accessory information corresponding to the accessory to be identified is greater than a preset threshold, then the candidate accessory corresponding to the candidate accessory image is taken as the target accessory corresponding to the image information of the accessory to be identified.

[0041] This step uses cross-modal semantic verification to determine the matching validity between candidate accessories and the accessory to be identified, ultimately determining the target accessory. Figure 2 The core verification step of "VLM vectorized retrieval (part number matching, image analysis)" in China.

[0042] In one implementation, the degree of matching between the accessory information of the candidate accessory corresponding to the candidate accessory image and the accessory information corresponding to the accessory to be identified is determined by the following method: The accessory information corresponding to the candidate accessory image, the accessory information corresponding to the accessory to be identified, and the guided recognition instructions constructed for accessory recognition requirements are input into the visual-language fusion model. The visual-language fusion model refers to a model with cross-modal semantic understanding capabilities of images and text, which can fuse and analyze the input image information and text information, and output matching results and related evaluation data.

[0043] The visual-language fusion model analyzes the matching degree of the input information to obtain the matching degree between the accessory information of the candidate accessory image and the accessory information of the accessory to be identified. Specifically, the visual-language fusion model extracts fields from the accessory information of the candidate accessory and the accessory information of the accessory to be identified, respectively extracting the field information corresponding to the key comparison fields from the accessory information of the candidate accessory and the accessory information of the accessory to be identified. Key comparison fields refer to the core fields used to determine whether the candidate accessory and the accessory to be identified match, including accessory identification, specification parameters, category attributes, etc. For each key comparison field, the visual-language fusion model determines the text semantic similarity and attribute consistency between the accessory information of the candidate accessory and the accessory information of the accessory to be identified and the field information corresponding to the key comparison field, and uses the text semantic similarity and attribute consistency between the accessory information of the candidate accessory and the accessory information of the accessory to be identified and the field information corresponding to the key comparison field as the text semantic similarity and attribute consistency corresponding to the key comparison field. The visual-language fusion model calculates the degree of matching between the accessory information of the candidate accessory image and the accessory information of the accessory to be identified based on the text semantic similarity, attribute consistency and preset weight coefficients corresponding to each key comparison field.

[0044] Specifically, the degree of matching between the candidate accessory information corresponding to the candidate accessory image and the accessory information corresponding to the accessory to be identified is determined in the following way: First, the accessory information of the candidate accessory corresponding to the candidate accessory image, the accessory information corresponding to the accessory to be identified, and the guiding identification instructions constructed for accessory identification needs (such as "identify the degree of matching between the candidate accessory and the accessory to be identified, focusing on comparing part numbers, specifications, and category attributes") are jointly input into the visual-language fusion model (i.e., Figure 2 The "VLM vectorized retrieval" mentioned in the text refers to the Visual-Language Model (VLM), which is the same concept and refers to a model with cross-modal semantic understanding capabilities. Subsequently, the visual-language fusion model is used to analyze the matching degree of the input information. The specific process is as follows: First, the visual-language fusion model extracts fields from the accessory information of the candidate accessory and the accessory information to be identified, extracting the field information corresponding to key comparison fields (such as part number, specification parameters, category attributes, and component name) to ensure the comprehensiveness of the comparison dimensions. Second, for each key comparison field, the visual-language fusion model calculates the textual semantic similarity (such as the character matching degree of the part number, the numerical similarity of the specification parameters, and the semantic relevance of the name) and attribute consistency (such as whether the category attributes are completely consistent and whether the component functions match) between the corresponding field information in the accessory information of the candidate accessory and the accessory information to be identified. The calculation results are used as the textual semantic similarity and attribute consistency corresponding to the key comparison field. The third step involves the visual-language fusion model calculating the weighted sum of the text semantic similarity and attribute consistency of each field based on the preset weight coefficients corresponding to each key comparison field (e.g., the part number has the highest weight, followed by the specification parameters, and then the category attributes), to obtain the final matching degree (the value ranges from 0 to 1, with the value closer to 1 indicating a higher matching degree).

[0045] After obtaining the matching degree, it is compared with a preset threshold (the preset threshold can be set according to the industry's requirements for recognition accuracy, such as 0.7-0.9): if the matching degree is greater than the preset threshold, it is determined that the candidate accessory and the accessory to be identified are successfully matched. The candidate accessory corresponding to the candidate accessory image is taken as the target accessory corresponding to the image information of the accessory to be identified, and the relevant information of the target accessory (such as accessory name, number, brand and model, and associated document references) is returned to the user through the interactive interface. Figure 2 In the "Return Parts Information, Image Standard" step, if the matching degree does not reach the preset threshold, the following data supplementation and incremental expansion process will be initiated.

[0046] In one implementation, to continuously enrich the data in the electronic catalog of parts and the OSS image library and improve the system's recognition coverage and accuracy, the method further includes the following step (this step is an optimization mechanism of the present invention, corresponding to...). Figure 2 The closed-loop process of "guided data entry" and "data entry (data vectorization)" in China: If the matching degree between the accessory information of the candidate accessory corresponding to the candidate accessory image and the accessory information corresponding to the accessory to be identified does not reach the preset threshold, the image features, extracted accessory information and preliminary classification label of the accessory to be identified are recorded, a data supplementation guidance instruction is generated and pushed to the user interaction interface, guiding the user to upload the relevant explanatory documents and complete information of the accessory to be identified; The system receives supplementary data uploaded by users, performs standardized processing on the supplementary data, and then adds it to the preset electronic catalog of accessories and the preset image library of object storage service, thereby achieving incremental expansion of data resources.

[0047] This can be understood as follows: if the matching degree between the candidate accessory information corresponding to the candidate accessory image and the accessory information corresponding to the accessory to be identified does not reach the preset threshold, it indicates that there is currently no accurate matching data for the accessory to be identified in the current electronic accessory catalog and OSS image library. At this time, the system automatically records the image features of the accessory to be identified (such as high-dimensional feature embedding vectors), the extracted accessory information, and the preliminary classification labels (such as the approximate category determined based on appearance features), and generates a data supplement guidance instruction (such as "No accurately matching accessory was found, please upload the accessory's documentation and complete information to improve the database"), pushes it to the user interface, and guides the user to upload the relevant documentation (such as product manuals, factory certificates of conformity) and complete information (such as accurate part numbers, specifications, and compatible vehicle models) for the accessory to be identified. After the user uploads the supplementary data, the system receives the supplementary data and performs unified standardization processing on the supplementary data according to the standardization processing rules of the electronic accessory catalog in S103. Subsequently, the processed standardized data is added to the preset electronic accessory catalog and the preset object storage service image library, respectively, to realize incremental expansion of data resources. At the same time, the system associates and stores the supplementary data with the corresponding image features and information of the parts to be identified, providing data support for the identification of similar parts in the future, forming a closed-loop mechanism of "data missing - user supplementation - database update - identification optimization".

[0048] In addition, after the target part is determined, the system also supports user feedback optimization: users can confirm or modify the returned target part information. Based on the user's feedback, the system adjusts parameters such as the weight coefficient of key comparison fields, preset thresholds, and similarity calculation rules, and records the data processing details, matching results, and user feedback during this recognition process. These are used as training samples for algorithm models such as vision-language fusion models and convolutional neural network encoders to continuously optimize the model's recognition performance and further improve the accuracy and reliability of subsequent part recognition.

[0049] As can be seen from the above technical solution, the method provided by the present invention first acquires the image information of the accessory to be identified, then extracts information from the image information to obtain the corresponding accessory information, and then uses the accessory information to query a preset electronic catalog of accessories. Based on the query results, it flexibly chooses whether to determine candidate accessory images from a preset object storage service image library, and then determines the corresponding accessory information based on the candidate accessory images and determines the target accessory by judging the degree of matching. This derivation process enables the method to fully combine the advantages of accurate querying of the electronic catalog of accessories with the visual matching capabilities of the object storage service image library, effectively making up for the shortcomings of existing single recognition schemes; wherein, by extracting information from the image information This system can comprehensively uncover key information related to parts, providing sufficient basis for subsequent matching and reducing misjudgments caused by missing information. By first querying the electronic parts catalog and then selectively using the image library for matching, it ensures recognition efficiency when there is clear matching information, and solves the recognition problem in scenarios with no matching or multiple matching, reducing the mismatch rate of similar-looking parts. Through quantitative judgment of the degree of matching, the accuracy of the recognition results is further improved, ultimately achieving high efficiency and precision in parts identification. This meets the automotive aftermarket's demand for high reliability and efficiency in parts identification, while reducing the time cost for users to locate parts information and improving the transaction efficiency and timeliness of data updates in related businesses.

[0050] like Figure 3 The image shows a specific embodiment of the accessory identification device described in this application. The device described in this embodiment is a physical device used to perform the method described in the above embodiments. Its technical solution is essentially the same as that of the above embodiments, and the corresponding descriptions in the above embodiments are also applicable to this embodiment. In this embodiment, the device includes: Information acquisition unit 301 is used to acquire image information of the accessory to be identified; Information extraction unit 302 is used to extract information from the image information of the accessory to be identified, and obtain accessory information corresponding to the accessory to be identified; The result query unit 303 is used to query a preset electronic catalog of accessories using the accessory information corresponding to the accessory to be identified, and obtain the query result. The image determination unit 304 is used to determine, from a preset object storage service image library, a candidate accessory image that matches the image information of the accessory to be identified if the number of candidate accessories that match the accessory information corresponding to the accessory to be identified is 0, equal to or greater than 2 in the query result. Information determination unit 305 is used to determine the accessory information of the candidate accessory corresponding to the candidate accessory image based on the candidate accessory image; The accessory determination unit 306 is used to determine the candidate accessory corresponding to the candidate accessory image as the target accessory corresponding to the image information of the accessory to be identified if the degree of matching between the accessory information of the candidate accessory corresponding to the candidate accessory image and the accessory information corresponding to the accessory to be identified is greater than a preset threshold.

[0051] Optionally, acquiring the image information of the accessory to be identified includes: Receive data related to the accessory to be identified uploaded by the user through the interactive interface. The data related to the accessory to be identified includes image files or document files containing images of the accessory. The data related to the component to be identified is parsed to extract the component image information; wherein, the component image information includes images of the component to be identified from different angles and disassembled explosion images of the component; The accessory image information is used as the image information of the accessory to be identified.

[0052] Optionally, the step of extracting information from the image information of the accessory to be identified to obtain accessory information corresponding to the accessory to be identified includes: An image region segmentation algorithm is used to segment the image information of the accessory to be identified into blocks, and the target image region containing the accessory identification text is located and extracted; The text information in the target image area is recognized and collected using an optical character recognition module to extract accessory attribute information; wherein, the accessory attribute information includes accessory identification, specification parameters and accessory description text information; The image information of the accessory to be identified is subjected to structured data detection. If structured data in tabular form is detected, a preset table recognition model is called to parse the field content in the table and convert the parsed field data into structured information in a unified format. The extracted accessory attribute information is associated and bound with preset standard fields to form accessory information corresponding to the accessory to be identified.

[0053] Optionally, the method for generating the preset electronic catalog of accessories includes the following steps: Collect structured data of multi-source components; The description specifications, coding systems and language formats of data from different sources are standardized to obtain standardized data. The processed standardized data is stored on a local storage medium or in a cloud storage system as a preset electronic catalog of accessories.

[0054] Optionally, determining candidate accessory images that match the image information of the accessory to be identified from a preset object storage service image library includes: Send a retrieval request to the preset object storage service image library to retrieve the set of accessory image samples stored in the object storage service image library; The image information of the accessory to be identified is subjected to feature enhancement preprocessing to obtain the preprocessed image information of the accessory to be identified; the feature enhancement preprocessing is a processing method that removes interference information in the image and highlights the key features of the accessory; A convolutional neural network encoder is used to extract features from the preprocessed image information of the accessory to be identified, thereby obtaining a high-dimensional feature embedding vector corresponding to the image information of the preprocessed accessory to be identified; Calculate the cosine similarity between the high-dimensional feature embedding vector and the feature vectors corresponding to each sample image in the accessory image sample set; The cosine similarity of each sample image is sorted from high to low, and the top sample images with the highest cosine similarity are selected as candidate accessory images that match the image information of the accessory to be identified.

[0055] Optionally, determining the accessory information of the candidate accessory corresponding to the candidate accessory image based on the candidate accessory image includes: Image recognition and semantic analysis are performed on the candidate component images to obtain the component information of the candidate components corresponding to the candidate component images; wherein, the component information of the candidate components corresponding to the candidate component images includes the category attributes, component names and key parameter information of the candidate components.

[0056] Optionally, the degree of matching between the accessory information of the candidate accessory corresponding to the candidate accessory image and the accessory information corresponding to the accessory to be identified is determined in the following way: The accessory information of the candidate accessory corresponding to the candidate accessory image, the accessory information corresponding to the accessory to be identified, and the guided recognition instructions constructed for accessory recognition requirements are input into the visual-language fusion model; The visual-language fusion model is used to analyze the matching degree of the input information to obtain the matching degree between the accessory information of the candidate accessory image and the accessory information of the accessory to be identified.

[0057] Optionally, the step of performing a matching degree analysis on the input information through the vision-language fusion model to obtain the matching degree between the accessory information of the candidate accessory image and the accessory information corresponding to the accessory to be identified includes: The visual-language fusion model extracts fields from the accessory information of the candidate accessory and the accessory information to be identified, respectively extracting the field information corresponding to the key comparison fields from the accessory information of the candidate accessory and the accessory information to be identified; The vision-language fusion model determines the text semantic similarity and attribute consistency between the accessory information of the candidate accessory and the accessory information of the accessory to be identified and the field information corresponding to the key comparison field for each key comparison field. The text semantic similarity and attribute consistency between the accessory information of the candidate accessory and the accessory information of the accessory to be identified and the field information corresponding to the key comparison field are used as the text semantic similarity and attribute consistency corresponding to the key comparison field. The visual-language fusion model calculates the degree of matching between the accessory information of the candidate accessory image and the accessory information of the accessory to be identified based on the text semantic similarity, attribute consistency and preset weight coefficients corresponding to each key comparison field.

[0058] Optionally, the device further includes a data expansion unit for: If the matching degree between the accessory information of the candidate accessory corresponding to the candidate accessory image and the accessory information corresponding to the accessory to be identified does not reach the preset threshold, the image features, extracted accessory information and preliminary classification label of the accessory to be identified are recorded, a data supplementation guidance instruction is generated and pushed to the user interaction interface, guiding the user to upload the relevant explanatory documents and complete information of the accessory to be identified; The system receives supplementary data uploaded by users, performs standardized processing on the supplementary data, and then adds it to the preset electronic catalog of accessories and the preset image library of object storage service, thereby achieving incremental expansion of data resources.

[0059] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. At the hardware level, the electronic device includes a processor, and optionally also includes an internal bus, a network interface, and a memory. The memory may include main memory, such as high-speed random-access memory (RAM), or it may also include non-volatile memory, such as at least one disk storage device. Of course, the electronic device may also include other hardware required for other services.

[0060] The processor, network interface, and memory can be interconnected via an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. This bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 4 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0061] Memory is used to store instructions for execution. Specifically, instructions for execution are computer programs that can be executed. Memory can include main memory and non-volatile memory, and it provides the processor with execution instructions and data.

[0062] In one possible implementation, the processor reads the corresponding execution instructions from non-volatile memory into main memory and then executes them. Alternatively, it may obtain the corresponding execution instructions from other devices to form an accessory identification device at the logical level. The processor executes the execution instructions stored in the memory to implement the accessory identification method provided in any embodiment of the present invention through the executed instructions.

[0063] The above is as described in the present invention. Figure 1 The accessory identification device provided in the illustrated embodiment executes a method that can be applied to or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by integrated logic circuits in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor.

[0064] The steps of the method disclosed in the embodiments of this invention can be directly manifested as being executed by a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method.

[0065] This invention also proposes a readable medium that stores execution instructions. When the stored execution instructions are executed by the processor of an electronic device, the electronic device can execute the accessory identification method provided in any embodiment of this invention, and specifically execute the method described above for data query.

[0066] The electronic devices described in the foregoing embodiments may be computers.

[0067] Those skilled in the art will understand that embodiments of the present invention can be provided as methods or computer program products. Therefore, the present invention can be implemented in a completely hardware embodiment, a completely software embodiment, or a combination of software and hardware.

[0068] The various embodiments in this invention are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0069] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0070] The above description is merely an embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A method for identifying accessories, characterized in that, The method includes: Obtain image information of the accessory to be identified; Information is extracted from the image information of the accessory to be identified to obtain the accessory information corresponding to the accessory to be identified; Using the accessory information corresponding to the accessory to be identified, a preset electronic catalog of accessories is queried to obtain the query results; If the number of candidate accessories matching the accessory information corresponding to the accessory to be identified is 0, equal to or greater than 2, then a candidate accessory image matching the image information of the accessory to be identified is determined from the preset object storage service image library. Based on the candidate accessory image, determine the accessory information of the candidate accessory corresponding to the candidate accessory image; If the matching degree between the accessory information of the candidate accessory corresponding to the candidate accessory image and the accessory information corresponding to the accessory to be identified is greater than a preset threshold, then the candidate accessory corresponding to the candidate accessory image is taken as the target accessory corresponding to the image information of the accessory to be identified.

2. The accessory identification method according to claim 1, characterized in that, The acquisition of image information of the accessory to be identified includes: Receive data related to the accessory to be identified uploaded by the user through the interactive interface. The data related to the accessory to be identified includes image files or document files containing images of the accessory. The data related to the component to be identified is parsed to extract the component image information; wherein, the component image information includes images of the component to be identified from different angles and disassembled explosion images of the component; The accessory image information is used as the image information of the accessory to be identified.

3. The accessory identification method according to claim 1, characterized in that, The step of extracting information from the image information of the accessory to be identified to obtain the accessory information corresponding to the accessory to be identified includes: An image region segmentation algorithm is used to segment the image information of the accessory to be identified into blocks, and the target image region containing the accessory identification text is located and extracted; The text information in the target image area is recognized and collected using an optical character recognition module to extract accessory attribute information; wherein, the accessory attribute information includes accessory identification, specification parameters and accessory description text information; The image information of the accessory to be identified is subjected to structured data detection. If structured data in tabular form is detected, a preset table recognition model is called to parse the field content in the table and convert the parsed field data into structured information in a unified format. The extracted accessory attribute information is associated and bound with preset standard fields to form accessory information corresponding to the accessory to be identified.

4. The accessory identification method according to claim 1, characterized in that, The method for generating the preset electronic catalog of accessories includes the following steps: Collect structured data of multi-source components; The description specifications, coding systems and language formats of data from different sources are standardized to obtain standardized data. The processed standardized data is stored on a local storage medium or in a cloud storage system as a preset electronic catalog of accessories.

5. The accessory identification method according to claim 1, characterized in that, The step of determining candidate accessory images that match the image information of the accessory to be identified from a preset object storage service image library includes: Send a retrieval request to the preset object storage service image library to retrieve the set of accessory image samples stored in the object storage service image library; The image information of the accessory to be identified is subjected to feature enhancement preprocessing to obtain the preprocessed image information of the accessory to be identified; the feature enhancement preprocessing is a processing method that removes interference information in the image and highlights the key features of the accessory; A convolutional neural network encoder is used to extract features from the preprocessed image information of the accessory to be identified, thereby obtaining a high-dimensional feature embedding vector corresponding to the image information of the preprocessed accessory to be identified; Calculate the cosine similarity between the high-dimensional feature embedding vector and the feature vectors corresponding to each sample image in the accessory image sample set; The cosine similarity of each sample image is sorted from high to low, and the top sample images with the highest cosine similarity are selected as candidate accessory images that match the image information of the accessory to be identified.

6. The accessory identification method according to claim 1, characterized in that, The step of determining the accessory information of the candidate accessory corresponding to the candidate accessory image based on the candidate accessory image includes: Image recognition and semantic analysis are performed on the candidate component images to obtain the component information of the candidate components corresponding to the candidate component images; wherein, the component information of the candidate components corresponding to the candidate component images includes the category attributes, component names and key parameter information of the candidate components.

7. The accessory identification method according to claim 1, characterized in that, The degree of matching between the accessory information of the candidate accessory image and the accessory information of the accessory to be identified is determined by the following method: The accessory information of the candidate accessory corresponding to the candidate accessory image, the accessory information corresponding to the accessory to be identified, and the guiding recognition instructions constructed for accessory recognition requirements are input into the vision-language fusion model; The visual-language fusion model is used to analyze the matching degree of the input information to obtain the matching degree between the accessory information of the candidate accessory image and the accessory information of the accessory to be identified.

8. The accessory identification method according to claim 7, characterized in that, The step of performing a matching degree analysis on the input information using the vision-language fusion model to obtain the matching degree between the accessory information of the candidate accessory image and the accessory information of the accessory to be identified includes: The visual-language fusion model extracts fields from the accessory information of the candidate accessory and the accessory information to be identified, respectively extracting the field information corresponding to the key comparison fields from the accessory information of the candidate accessory and the accessory information to be identified; The vision-language fusion model determines the text semantic similarity and attribute consistency between the accessory information of the candidate accessory and the accessory information of the accessory to be identified and the field information corresponding to the key comparison field for each key comparison field. The text semantic similarity and attribute consistency between the accessory information of the candidate accessory and the accessory information of the accessory to be identified and the field information corresponding to the key comparison field are used as the text semantic similarity and attribute consistency corresponding to the key comparison field. The visual-language fusion model calculates the degree of matching between the accessory information of the candidate accessory image and the accessory information of the accessory to be identified based on the text semantic similarity, attribute consistency and preset weight coefficients corresponding to each key comparison field.

9. The accessory identification method according to claim 1, characterized in that, The method also includes If the matching degree between the accessory information of the candidate accessory corresponding to the candidate accessory image and the accessory information corresponding to the accessory to be identified does not reach the preset threshold, the image features, extracted accessory information and preliminary classification label of the accessory to be identified are recorded, a data supplementation guidance instruction is generated and pushed to the user interaction interface, guiding the user to upload the relevant explanatory documents and complete information of the accessory to be identified; The system receives supplementary data uploaded by users, performs standardized processing on the supplementary data, and then adds it to the preset electronic catalog of accessories and the preset image library of object storage service, thereby achieving incremental expansion of data resources.

10. An accessory identification device, characterized in that, The device includes: The information acquisition unit is used to acquire image information of the accessory to be identified; An information extraction unit is used to extract information from the image information of the accessory to be identified, and obtain accessory information corresponding to the accessory to be identified; The result query unit is used to query a preset electronic catalog of accessories using the accessory information corresponding to the accessory to be identified, and obtain the query result. An image determination unit is used to determine, if the number of candidate accessories matching the accessory information corresponding to the accessory to be identified is 0, equal to or greater than 2, from a preset object storage service image library; An information determination unit is used to determine the accessory information of the candidate accessory corresponding to the candidate accessory image based on the candidate accessory image; The accessory determination unit is configured to, if the degree of matching between the accessory information of the candidate accessory corresponding to the candidate accessory image and the accessory information corresponding to the accessory to be identified is greater than a preset threshold, then use the candidate accessory corresponding to the candidate accessory image as the target accessory corresponding to the image information of the accessory to be identified.

Images