An AI-based textile industry price market collection method and system
By using AI technology to process non-standardized table and price quotation images in the textile industry, automated data collection and unified format generation have been achieved, solving the problems of high cost and low accuracy in existing technologies and improving the efficiency and accuracy of data collection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANGHAI ZHIJING INFORMATION TECH CO LTD
- Filing Date
- 2026-02-27
- Publication Date
- 2026-06-09
Smart Images
Figure CN122176734A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of artificial intelligence, image recognition and natural language processing, specifically to an AI-based method and system for collecting price information in the textile industry. Background Technology
[0002] In the textile industry, yarn and other product manufacturers are the primary providers of price information. These factories typically publish daily product price quotes in tabular format via social media channels like WeChat. To quickly extract price data from these images, the industry commonly uses OCR (Optical Character Recognition) technology, but this method has significant technical limitations in practical applications.
[0003] Currently, there is no unified standard for creating price lists in the textile industry. Price lists from different factories vary greatly in format, number of columns, header settings, and layout. Furthermore, many price lists use complex table-within-table structures to present price data for different product categories. Conventional OCR technology can only achieve basic table outline recognition and text extraction, lacking artificial intelligence-based intelligent analysis and processing capabilities. When processing table-within-table images, it cannot accurately identify sub-table boundaries, and the returned data can only be presented in a raw, write-back format, unable to be reorganized according to actual application needs.
[0004] In practical applications of price information collection in the textile industry, it is necessary to extract the various sub-tables in a queue manner and ultimately integrate them into a large table or master table with a unified format for storage and analysis. The output of conventional OCR recognition technology cannot directly meet this requirement, resulting in a large amount of manual secondary processing for the analysis, storage, and application of the collected data, or the development of separate adaptation programs for price quotation tables of different factories. This not only significantly increases the development, maintenance, and time costs of price information collection, but also easily leads to a high error rate in data processing due to manual intervention, affecting the accuracy and timeliness of price information data. Summary of the Invention
[0005] The purpose of this invention is to provide an AI-based method and system for collecting price information in the textile industry, aiming to improve the problems of existing OCR table recognition technology in processing price quote images in the textile industry, such as the inability to output data in a uniform format and low recognition efficiency and accuracy.
[0006] This invention is implemented as follows: According to a first aspect of the present invention, the present invention provides an AI-based method for collecting price information in the textile industry, comprising the following steps: S1. Image Repair and Cleanup: Process the tabular price quotation images of textile factories, remove irrelevant information outside the table, and retain only the table area; S2, Table Cropping: Automatically crop the repaired and cleaned images according to a table, dividing the images into multiple smallest sub-table images without a table in the table; S3, Sub-table OCR Recognition: Perform OCR recognition on the smallest sub-table image, extract the row, column, and header names of the sub-table, as well as the corresponding text and / or numeric data, and return the sub-table data containing complete structured information; S4. Header Row to Column Conversion: Convert the format of each sub-table data after OCR recognition, extract the header row information and convert it into column information, generate an intermediate table with a uniform format, and realize the one-to-one association between sub-table data and header; S5. Multi-table concatenation and integration: Align the columns of all the intermediate tables corresponding to the sub-tables according to the principle of the same attributes, fill in the missing values of the columns, and then concatenate and integrate the multiple intermediate tables into a large table with a unified format in a queue manner.
[0007] Furthermore, in step S1, an AI visual processing algorithm is used to repair and clean the image, while an AI image enhancement algorithm is used to repair the blurry and broken lines in the table, strengthen the table outline features, and improve the image recognition rate.
[0008] Furthermore, the AI image enhancement algorithm is an image restoration algorithm based on convolutional neural networks, which specifically repairs the blurry and broken features of table lines and enhances the edge and contour information of the table.
[0009] Furthermore, in step S2, the smallest sub-table image is an independent table image that cannot be further split into tables, and each smallest sub-table image contains only a single atomic-level table unit.
[0010] Furthermore, in step S4, the header of the sub-table is used as the new column name, and the row data of the sub-table and the header form a one-to-one correspondence, thereby achieving the format unification of non-standardized sub-tables.
[0011] According to a second aspect of the present invention, the present invention provides an AI-based price information collection system for the textile industry, used to implement the aforementioned AI-based price information collection method for the textile industry, including... The image receiving module is used to receive quotation images in tabular form and transmit them to the AI preprocessing module; The AI preprocessing module has built-in AI visual processing algorithms and AI image enhancement algorithms, which are used to perform image repair and cleaning steps; The intelligent cutting module is equipped with an AI target detection and contour recognition model trained on samples from the textile industry to perform the table-by-table cutting steps; The OCR recognition module deploys an OCR recognition model trained with textile industry features and executes sub-table OCR recognition steps. The format conversion module has a built-in AI data format conversion model and performs the header row to column conversion step; The data integration module, equipped with an AI data fusion and alignment model, performs multi-table concatenation and integration steps; Furthermore, it also includes an AI model optimization module, which is used to continuously iterate and optimize the parameters of each model based on the newly added quotation form format samples in the textile industry, so as to improve the adaptability to new non-standard quotation forms.
[0012] Furthermore, the AI model optimization module establishes a bidirectional data connection with the AI preprocessing module. Based on the newly added price list image samples and table line feature samples from the textile industry, the AI model optimization module continuously iterates and optimizes the AI visual processing algorithm and the AI image enhancement algorithm based on convolutional neural networks in the AI preprocessing module to improve the accuracy and efficiency of image restoration and cleaning, and adapt to the processing needs of new non-standardized price lists.
[0013] Furthermore, it also includes a data output module and a database module. The data output module is used to store the large table data in a unified format output by the data integration module and write it into the database module.
[0014] Furthermore, the data output module includes a format conversion unit, used to convert the unified format large table data output by the data integration module into commonly used format files, which include at least Excel files, CSV files, JSON files and XML files.
[0015] Compared with existing technologies, the advantages of this invention are as follows: This invention forms a fully automated solution encompassing the entire process from receiving and cleaning up quotation images, cropping tables within tables, identifying sub-tables, format conversion, multi-table integration, data output, and model optimization. Through five core steps, relying on AI visual processing, image enhancement, and OCR recognition technologies, this invention achieves accurate processing of quotation images from non-standardized tables and the generation of large tables with a unified format. Furthermore, this invention continuously iterates and optimizes through an AI model optimization module, improving its adaptability to new non-standardized quotation tables. Compared with existing technologies, this invention eliminates the need for manual secondary processing, significantly reducing the cost of price information collection and substantially improving the efficiency, accuracy, and timeliness of data collection. Attached Figure Description
[0016] Figure 1 This is a flowchart of the AI-based price information collection method for the textile industry provided by the present invention. Figure 2 This is a block diagram of the module structure of the AI-based textile industry price information collection system provided by the present invention; Figure 3This is a schematic diagram of a table-style quotation image before processing; Figure 4 This is a table-style price quote image that has been repaired and cleaned. Figure 5 This is a table-format price quote image after cropping the table. Figure 6 This is a schematic diagram of the result after OCR recognition of the sub-table; Figure 7 This is a schematic diagram of the result of a sub-table after the header row has been transformed into a column; Figure 8 This is a schematic diagram of the final result of a large table with a unified format after multiple tables have been linked and integrated. Detailed Implementation
[0017] The following description, in conjunction with the accompanying drawings and specific embodiments, provides further details: Example 1
[0018] This embodiment provides an AI-based method for collecting price information in the textile industry, such as... Figure 1 As shown, it includes the following steps: S1. Image Repair and Cleanup: This process removes irrelevant information outside the table from the tabular price quote images from textile factories, retaining only the table area. The original tabular price quote image from the textile factory is shown below. Figure 3 As shown, irrelevant information typically includes redundant content such as factory signatures, quotation dates, and non-table remarks. After cleaning, only the pure table area shown in the image is retained, such as... Figure 4 As shown, the original quotation image, after image restoration and cleaning, lays a solid foundation for subsequent steps such as table segmentation and data recognition. Simultaneously, to further improve the image recognition accuracy of subsequent steps, this step employs AI visual processing algorithms for efficient cleaning, combined with AI image enhancement algorithms to optimize the table structure. It specifically repairs issues such as blurred, broken, and faded lines caused by factors like shooting angle deviations, insufficient lighting, and losses during transmission. The AI image enhancement algorithm used is an image restoration algorithm based on convolutional neural networks. This algorithm can accurately capture the blurred and broken features of table lines, perform targeted repair, effectively strengthen the table's edge and contour information, and further improve the accuracy of subsequent sub-table cropping and OCR recognition.
[0019] S2. Table Cropping: After image repair and cleaning, crop the repaired image according to a table, dividing the image into multiple smallest sub-table images without a table in the table, such as... Figure 5As shown, the smallest sub-table image is an independent table image that cannot be further divided. Each smallest sub-table image contains only a single atomic-level table unit, which is a basic table unit that cannot be further divided in terms of both physical borders and logical hierarchy. These sub-tables may contain multiple rows and columns of data, but they do not contain any form of nested tables (i.e., table-within-a-table structure). Each smallest sub-table has a complete header and data area, serving as the basic processing unit for subsequent OCR recognition and data integration. This splitting method simplifies the originally complex table structure into multiple independent, simple sub-tables, significantly reducing the complexity of subsequent OCR recognition and facilitating the accurate extraction of complete data information from each sub-table.
[0020] S3. Sub-table OCR Recognition: Each cropped sub-table image is processed using OCR to extract the row count, column count, header name, and corresponding text, numeric, or mixed text and numeric data within each cell. After recognition, the sub-table data containing complete structured information is returned. The processing result is as follows: Figure 6 As shown. This ensures that subsequent format conversion and data integration steps can proceed smoothly, avoiding impact on the overall processing results due to missing data or recognition errors.
[0021] S4. Header Row to Column Conversion: The format of each sub-table data after OCR recognition is converted, the header row information is extracted and converted into column information, generating a uniformly formatted intermediate table. This establishes a one-to-one association between the sub-table data and the header. The processing result is as follows: Figure 7 As shown. The specific operation involves extracting the header row information of each sub-table, converting the originally horizontally arranged header rows into vertical column names, and simultaneously establishing a one-to-one correspondence between each row of data in the sub-table and its corresponding column name. Through this conversion method, the format of non-standardized sub-tables with different formats, header settings, and column numbers can be unified, generating a uniform intermediate table, which fully prepares for the subsequent multi-table concatenation and integration steps.
[0022] S5. Multi-table concatenation and integration: All intermediate tables corresponding to the sub-tables are aligned according to the principle of identical attributes, and missing columns are padded with null values. Then, multiple intermediate tables are concatenated and integrated into a single large table with a unified format, facilitating subsequent data storage, querying, analysis, and application. Figure 8 As shown. Example 2
[0023] This embodiment provides an AI-based price information collection system for the textile industry, used to implement the AI-based price information collection method for the textile industry provided in Embodiment 1. Figure 2As shown, the system includes multiple modules that work together, such as an image receiving module, an AI preprocessing module, an intelligent cropping module, an OCR recognition module, a format conversion module, and a data integration module, to collect and process price information for the textile industry.
[0024] The image receiving module, serving as the system's input, primarily receives table-format price quote images sent by textile factories via various common channels such as WeChat, web uploads, and local uploads. Upon receipt, it transmits these images in real-time to the AI preprocessing module, ensuring the timely entry of the price quote images into subsequent processing and maximizing data collection efficiency. The AI preprocessing module is the core module for executing the image restoration and cleaning steps in Example 1. It incorporates AI visual processing algorithms and AI image enhancement algorithms, strictly adhering to the image restoration and cleaning requirements in Example 1. It accurately cleans irrelevant information outside the table, repairs table lines, and enhances the outline. After processing, the cleaned and restored table image is precisely transmitted to the intelligent cropping module.
[0025] The intelligent cropping module is equipped with an AI object detection and contour recognition model trained on a large number of price list samples from the textile industry. This model can accurately identify the table segmentation boundaries and table outlines in the repaired table image. Strictly following the requirements of the table cropping steps in Example 1, it automatically crops the image, dividing it into multiple smallest sub-table images without table segments. All cropped sub-table images are then transmitted one by one to the OCR recognition module. The OCR recognition module internally deploys an OCR recognition model trained on textile industry features. This model has been specifically optimized for features such as text, numbers, and professional specification symbols in textile industry price lists. It can accurately execute the sub-table OCR recognition steps in Example 1, extracting the row, column, and header names and corresponding data of the sub-table, generating complete structured sub-table data, and transmitting it to the format conversion module.
[0026] The format conversion module incorporates an AI data format conversion model, primarily used to execute the header row to column conversion step in Example 1. After receiving the sub-table data transmitted by the OCR recognition module, it performs format standardization conversion, extracts the header row information and converts it into column names, generating a uniformly formatted intermediate table. All intermediate tables are then transmitted to the data integration module. The data integration module is equipped with an AI data fusion and alignment model. Its core function is to execute the multi-table concatenation and integration step in Example 1, aligning attribute columns and filling in missing null values in all intermediate tables. Finally, it concatenates and integrates multiple intermediate tables into a single large table with a unified format, completing the data integration process.
[0027] like Figure 2As shown, the AI-based textile industry price information collection system also includes an AI model optimization module. This module is primarily used to continuously iterate and optimize the relevant models in each module based on newly added price list format samples from the textile industry. Specifically, the AI model optimization module establishes a bidirectional data connection with the AI preprocessing module. The AI model optimization module continuously iterates and optimizes the AI visual processing algorithm and the AI image enhancement algorithm based on convolutional neural networks in the AI preprocessing module based on newly added price list image samples and table line feature samples from the textile industry. This improves the accuracy and efficiency of image restoration and cleaning, ensuring the system can always adapt to the processing needs of new, non-standardized price lists, and extending the system's lifespan and applicability. In addition, the system also includes a data output module and a database module. The data output module is mainly used to store the large table data in a unified format output by the data integration module and write it into the database module to achieve long-term data preservation and subsequent retrieval. To adapt to the data usage needs of different scenarios in the textile industry, the data output module also includes a format conversion unit. This format conversion unit can convert the large table data in a unified format output by the data integration module into a variety of commonly used format files. These commonly used format files include at least Excel files, CSV files, JSON files, and XML files, which are convenient for different market analysis platforms and data management systems to directly call and use.
[0028] In summary, this invention forms a fully automated solution encompassing the entire process from receiving and cleaning price quote images, cropping tables within tables, identifying sub-tables, format conversion, multi-table integration, data output, and model optimization. Through five core steps, leveraging AI visual processing, image enhancement, and OCR recognition technologies, this invention achieves accurate processing of non-standardized price quote images and the generation of unified format tables. Furthermore, the invention utilizes an AI model optimization module for continuous iterative optimization, enhancing its adaptability to new non-standardized price quote tables. Compared to existing technologies, this invention eliminates the need for manual secondary processing, significantly reducing price information collection costs and substantially improving the efficiency, accuracy, and timeliness of data collection. It provides reliable technical support for the rapid collection, efficient analysis, and scientific decision-making of price information in the textile industry, demonstrating significant practicality and industry application value.
[0029] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. An AI-based textile industry price trend collection method, characterized by, Includes the following steps: S1. Image Repair and Cleanup: Process the tabular price quotation images of textile factories, remove irrelevant information outside the table, and retain only the table area; S2, Table Cropping: Crops the repaired and cleaned image according to a table, dividing the image into multiple smallest sub-table images without a table in the table; S3, Sub-table OCR Recognition: Perform OCR recognition on the smallest sub-table image, extract the row, column, and header names of the sub-table, as well as the corresponding text and / or numeric data, and return the sub-table data containing complete structured information; S4. Header Row to Column Conversion: Convert the format of each sub-table data after OCR recognition, extract the header row information and convert it into column information, generate an intermediate table with a uniform format, and realize the one-to-one association between sub-table data and header; S5. Multi-table concatenation and integration: Align the columns of all the intermediate tables corresponding to the sub-tables according to the principle of the same attributes, fill in the missing values of the columns, and then concatenate and integrate the multiple intermediate tables into a large table with a unified format in a queue manner.
2. The AI-based textile industry price trend collection method according to claim 1, characterized in that, In step S1, an AI visual processing algorithm is used to repair and clean the image, and an AI image enhancement algorithm is used to repair the blurry and broken lines in the table, strengthen the table outline features, and improve the image recognition rate. 3.The AI-based textile industry price trend collecting method of claim 2, wherein, The AI image enhancement algorithm is an image restoration algorithm based on convolutional neural networks. It targets the blurring and broken features of table lines to restore them and enhance the edge and contour information of the table.
4. The AI-based textile industry price trend collection method according to claim 1, characterized in that, In step S2, the smallest sub-table image is an independent table image that cannot be further split into tables, and each smallest sub-table image contains only a single atomic-level table unit. 5.The AI-based textile industry price trend collecting method according to claim 1, wherein, In step S4, the header of the sub-table is used as the new column name, and the row data of the sub-table and the header form a one-to-one correspondence, thereby achieving the format unification of non-standardized sub-tables.
6. An AI-based price information collection system for the textile industry, used to implement the AI-based price information collection method for the textile industry as described in any one of claims 1-5, characterized in that, include: The image receiving module is used to receive quotation images in tabular form and transmit them to the AI preprocessing module; The AI preprocessing module has built-in AI visual processing algorithms and AI image enhancement algorithms, which are used to perform image repair and cleaning steps; The intelligent cutting module is equipped with an AI target detection and contour recognition model trained on samples from the textile industry to perform the table-by-table cutting steps; The OCR recognition module deploys an OCR recognition model trained with textile industry features and executes sub-table OCR recognition steps. The format conversion module has a built-in AI data format conversion model and performs the header row to column conversion step; The data integration module, equipped with an AI data fusion and alignment model, performs the step of connecting and integrating multiple tables.
7. The AI-based price information collection system for the textile industry according to claim 6, characterized in that, It also includes an AI model optimization module, which is used to continuously iterate and optimize the parameters of each model based on the newly added quotation form format samples in the textile industry, so as to improve the adaptability to new non-standard quotation forms.
8. The AI-based price information collection system for the textile industry according to claim 7, characterized in that, The AI model optimization module establishes a bidirectional data connection with the AI preprocessing module. Based on the newly added price list image samples and table line feature samples from the textile industry, the AI model optimization module continuously iterates and optimizes the AI visual processing algorithm and the AI image enhancement algorithm based on convolutional neural networks in the AI preprocessing module to improve the accuracy and efficiency of image restoration and cleaning, and adapt to the processing needs of new non-standardized price lists.
9. The AI-based price information collection system for the textile industry according to claim 6, characterized in that, It also includes a data output module and a database module. The data output module is used to store the large table data in a unified format output by the data integration module and write it into the database module.
10. The AI-based price information collection system for the textile industry according to claim 9, characterized in that, The data output module includes a format conversion unit, which is used to convert the unified format large table data output by the data integration module into commonly used format files. The commonly used format files include at least Excel files, CSV files, JSON files, and XML files.