An AI-based supplier procurement management method

Abstract

This invention relates to the field of supply chain information management technology, and more particularly to an AI-based supplier procurement management method. The invention pre-acquires product images from product information, obtains contour recognition results from scaled product images, filters quality confidence feature images based on the overlap of contour recognition results, compares these quality confidence feature images with images in an existing product database to determine whether to store or delete the supplier's product information, determines the comprehensive characteristic indicators of the product corresponding to the product image in production, and determines whether to compress the product image into several data packets for transmission to the review terminal based on the comparison results of the comprehensive characteristic indicators and their reference values. This invention optimizes the information processing process in complex procurement scenarios for manufacturing enterprises, enabling targeted extraction of supplier information content and improving the scientific nature of supplier information management.

Description

Technical Field

[0001] This invention relates to the field of supply chain information management technology, and in particular to an AI-based supplier procurement management method. Background Technology

[0002] In traditional supplier procurement management, manual processing of supplier-provided product information is inefficient and prone to omissions or errors when dealing with large volumes of data. Product quality assessment relies heavily on human experience, leading to subjectivity and potentially allowing substandard products to enter the production process. Furthermore, the lack of unified digital management of product information makes it difficult to rationally determine procurement priorities. As AI technology is increasingly integrated into supply chain management, image recognition and big data analytics can improve efficiency and objectivity. Using AI algorithms to process and analyze product images enables highly accurate quality screening and information comparison.

[0003] For example, Chinese Patent Publication No. CN115907673A discloses a supply chain system including an AI platform, a business platform, and a client. The client is used by users to log in to the business platform and upload document data. The business platform receives the document data uploaded by users on the client and transmits the document data to the AI ​​platform. Based on the data fed back by the AI ​​platform, it performs document verification, deduplication, and generation to complete document signing. This includes a business system, a workflow system, a task scheduling system, an electronic contract system, a file storage system, a cache database, and a storage database. The AI ​​platform intelligently processes the document data transmitted from the business platform, converting it into structured data and image information and feeding it back to the business platform. This includes an optical character recognition system and an RPA robot. This solves the problems of tedious and repetitive manual identification and transcription, achieving accurate and efficient identification of document information.

[0004] The following problems still exist in the existing technology:

[0005] Existing technologies do not consider optimizing the information processing process for manufacturing enterprises in complex procurement scenarios, cannot eliminate the problem of insufficient information utilization efficiency in traditional procurement management, and cannot extract the information content of supplier information in a targeted manner, thus affecting the scientific nature of supplier information management. Summary of the Invention

[0006] To address this, the present invention provides an AI-based supplier procurement management method to overcome the problems of insufficient information utilization efficiency in traditional procurement management and the inability to extract targeted information content from supplier information in existing technologies.

[0007] To achieve the above objectives, this invention provides an AI-based supplier procurement management method, comprising:

[0008] The product information sent by the supplier is obtained in advance at the information receiving interface, and the product image is extracted from the product information;

[0009] Each product image is scaled based on a pre-set scaling factor, and the contour recognition results of the product images are obtained under different scaling states. Quality confidence feature images are then selected based on the overlap of the contour recognition results.

[0010] The quality confidence feature image is uploaded to a pre-built existing product database for comparison and analysis. Based on the results of the comparison and analysis, it is determined whether to store or delete the supplier's product information.

[0011] In response to the determination result of storing the product image, a first dimension feature index and a second dimension feature index of the product corresponding to the product image in production are determined, wherein the first dimension feature index is determined according to the time progress node of participating in the product production, and the second dimension feature index is determined according to the comparison result of product information and historical product information.

[0012] The comprehensive feature index of the product is determined based on the first dimension feature index and the second dimension feature index. Based on the comparison result of the comprehensive feature index and the comprehensive feature index reference value, it is determined whether to compress the product image into several data packets and transmit them to the review terminal.

[0013] Furthermore, the process of scaling each product image based on a pre-set scaling factor includes:

[0014] Pre-set several scaling factors;

[0015] The scaling dimensions of each product image are determined based on each scaling factor, and the scaling state image is determined according to the scaling dimensions.

[0016] The scaling factor values ​​are arranged in an arithmetic sequence.

[0017] Furthermore, the process of obtaining the contour recognition results of product images under different scaling states includes:

[0018] Determine the contour recognition lines within the image at each zoom level;

[0019] Place the contour recognition lines in each scaled image under the same scale, and calculate the overlap of any two contour recognition lines among a number of contour recognition lines.

[0020] The average value of the overlap between several sets of two contour recognition lines is determined as the quality confidence value of the product image.

[0021] Furthermore, the product image corresponding to the maximum value of the quality confidence characterization value is selected as the quality confidence feature image.

[0022] Furthermore, the process of determining whether to store or delete the supplier's product information based on the results of the comparative analysis includes:

[0023] Calculate the intersection-union ratio (IUR) of the quality confidence feature image with each sample image in the existing product database, and determine the maximum IUR value;

[0024] If the maximum value of the cross-union ratio is less than the preset cross-union ratio threshold, it is determined that the product information should be stored.

[0025] If the maximum value of the cross-union ratio is greater than or equal to the preset cross-union ratio threshold, it is determined that the product information should be cleared.

[0026] Furthermore, the process of determining the first dimension feature index of the product corresponding to the product image in production includes:

[0027] Determine the duration of the timeline between the start time of production and the timeline of production.

[0028] The value obtained by dividing the duration by the total duration of commodity production is determined as the first dimension feature index.

[0029] Furthermore, the process of determining the second-dimensional feature indicators of the product corresponding to the product image in production includes:

[0030] Extract several performance parameters from the product information;

[0031] The performance parameters of the product information are compared with the performance parameters corresponding to historical product information to determine the number of performance parameter groups with differences.

[0032] The ratio of the number of performance parameter groups with differences to the total number of performance parameter groups in historical product information is determined as the second dimension feature index.

[0033] Furthermore, the value obtained by dividing the second dimension feature index by the first dimension feature index is determined as the comprehensive feature index.

[0034] Furthermore, the process of determining whether to compress the product image into several data packets and transmit them to the review terminal includes:

[0035] The comprehensive feature index is compared with the preset comprehensive feature index reference value;

[0036] If the comprehensive feature index is greater than the comprehensive feature index reference value, then it is determined that the product image will be compressed into several data packets and transmitted to the review terminal.

[0037] Furthermore, the process of compressing the product image into several data packets and transmitting them to the review terminal includes:

[0038] Determine the outline recognition lines within the product image;

[0039] Product images where the contour recognition line intersects with the edge of the product image are compressed into the first data packet, and product images where the contour recognition line does not intersect with the edge of the product image are compressed into the second data packet;

[0040] The first data packet and the second data packet are transmitted to the auditing terminal in sequence.

[0041] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention pre-acquires product information sent by suppliers and extracts product images at the information receiving interface. By scaling each product image, the contour recognition results of the product images are obtained. Quality confidence feature images are selected based on the overlap of the contour recognition results. By comparing the quality confidence feature images with images in the existing product database, it is determined whether to store or delete the supplier's product information. The comprehensive characteristic indicators of the product corresponding to the product image in production are determined. Based on the comparison results of the comprehensive characteristic indicators and their reference values, it is determined whether to compress the product image into several data packets and transmit them to the review terminal. This invention optimizes the information processing process for manufacturing enterprises in complex procurement scenarios, enabling targeted extraction of supplier information content and improving the scientific nature of supplier information management.

[0042] Furthermore, the present invention pre-sets several scaling coefficients arranged in an arithmetic sequence, and determines the scaling length and width dimensions of each product image based on these coefficients, thereby generating images in different scaling states. This approach can capture the features of product images from multiple scaling dimensions, providing diverse image samples for contour recognition and overlap analysis under different scaling states.

[0043] Furthermore, this invention generates product images at different scaling states and extracts contour recognition lines. After conversion to the same scale, the overlap of any two contour lines is calculated and the average value is taken as the quality confidence characterization value. Finally, the image with the largest value is selected as the quality confidence feature image. This is because the high overlap of contour lines at multiple scaling states indicates that the image is less affected by shooting angle, lighting, etc., and can stably reflect the true geometric features of the product, thereby improving the reliability and efficiency of product information processing.

[0044] Furthermore, in the practical scenario of managing product information for manufacturing suppliers, this invention compares the previously screened quality confidence feature images with the existing sample images in the product database, calculates the ratio of the overlapping area of ​​the two contours to the total area (i.e., the intersection-union ratio), and finds the maximum value to measure the similarity between the new product and the existing products in the database. When the maximum value is lower than a preset threshold, it indicates that the new product is significantly different from the existing samples in the database and belongs to a new specification product. When the calculated maximum value of the intersection-union ratio between the quality confidence feature image and the sample images in the existing product database reaches or exceeds the preset threshold, it means that the product is highly similar to the existing samples in the database in terms of features, and there is a situation where the supplier's product image information is misusing other suppliers' information. In supply management, there is a phenomenon where the provided product is actually similar to other suppliers' existing products, but attempts to enter the procurement process with new information. Clearing such information can avoid memory occupation caused by information duplication, prevent redundancy in the existing product database from affecting the efficiency of subsequent information processing, and reduce the procurement risks that may be caused by information misuse. It realizes intelligent filtering of supplier product information, balances information integrity and management efficiency, and provides reliable data support for procurement decisions.

[0045] Furthermore, by determining comprehensive characteristic indicators, this invention avoids decision-making biases that prioritize low-risk components based solely on time urgency while neglecting high-risk and urgent components. It also allows for a direct differentiation of the risk and urgency of goods through numerical values, enabling procurement and review resources to be prioritized for key areas of focus. This ensures production progress while reducing compatibility risks caused by performance differences, thereby improving procurement decision-making efficiency and production coordination stability.

[0046] Furthermore, this invention divides a product image into a first data packet containing local key features and a second data packet containing the complete global shape based on whether the contour recognition line intersects with the image edge, and transmits them sequentially. This allows the review terminal to prioritize obtaining image information of the local key areas of the product, enabling rapid review of core quality features, while supplementing the overall shape verification through the subsequently transmitted global image. This ensures the integrity of the review while reducing the consumption of transmission bandwidth and review resources by non-critical information, thereby improving the efficiency and accuracy of product quality review. It is particularly suitable for industrial quality inspection scenarios that require both detail accuracy and overall compliance. Attached Figure Description

[0047] Figure 1 This invention provides a temperature and humidity control system for manufacturing welding area floors.

[0048] Figure 2 This is a step diagram illustrating the process of obtaining the contour recognition result of a product image according to an embodiment of the present invention;

[0049] Figure 3A flowchart illustrating the logic for determining whether to store or clear the supplier's product information;

[0050] Figure 4 A flowchart illustrating the steps for determining the second-dimensional feature index in an embodiment of the present invention. Detailed Implementation

[0051] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0052] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0053] It should be noted that in the description of this invention, the terms "upper," "lower," "inner," "outer," etc., which indicate the direction or positional relationship, are based on the direction or positional relationship shown in the drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0054] Please see Figure 1 The diagram illustrates the steps of an AI-based supplier procurement management method according to an embodiment of the present invention. The AI-based supplier procurement management method of the present invention includes:

[0055] Step S10: Obtain the product information sent by the supplier in advance at the information receiving interface, and extract the product image from the product information;

[0056] Specifically, the present invention can establish a data transmission channel between supplier information and the receiving system based on a standardized data format through an API interface or a data transmission protocol interface, which is well known to those skilled in the art and will not be described in detail here.

[0057] Step S20: Scale each product image based on a pre-set scaling factor, and obtain the contour recognition results of the product images under different scaling states. Select quality confidence feature images based on the overlap of the contour recognition results.

[0058] Step S30: Upload the quality confidence feature image to a pre-built existing product database for comparison and analysis, and determine whether to store or delete the supplier's product information based on the results of the comparison and analysis.

[0059] Specifically, the existing product database in this invention can be a database that integrates multi-dimensional content such as historical procurement data, product standard information, and quality benchmark characteristics. It includes the model, code, material, design parameters, and applicable scenarios of all historically procured products. It establishes a fast retrieval mechanism through product codes to ensure that newly uploaded feature images can be efficiently matched with corresponding standard images and historical data, thereby improving comparison efficiency.

[0060] Step S40: In response to the determination result of storing the product image, determine the first dimension feature index and the second dimension feature index of the product corresponding to the product image in production, wherein the first dimension feature index is determined according to the time progress node of participating in product production, and the second dimension feature index is determined according to the comparison result of product information and historical product information.

[0061] Step S50: Determine the comprehensive feature index of the product based on the first dimension feature index and the second dimension feature index, and determine whether to compress the product image into several data packets and transmit them to the review terminal based on the comparison result of the comprehensive feature index and the comprehensive feature index reference value.

[0062] Specifically, the verification terminal in this invention can be a computer terminal used to receive product information.

[0063] Specifically, the process of scaling each product image based on a pre-set scaling factor includes:

[0064] Pre-set several scaling factors;

[0065] The scaling dimensions of each product image are determined based on each scaling factor, and the scaling state image is determined according to the scaling dimensions.

[0066] The scaling factor values ​​are arranged in an arithmetic sequence.

[0067] For example, the scaling factor k can be set to 0.8, 0.6, 0.4, and 0.2 respectively. If the original length of the product image is L in pixels and the original width is W, then for each scaling factor k, the scaled length is L×k and the scaled width is W×k.

[0068] If the original size of a product image is 1000 pixels (length) × 800 pixels (width);

[0069] When the scaling factor is 0.8, the length after scaling is 1000 × 0.8 = 800, and the width is 800 × 0.8 = 640.

[0070] When the scaling factor is 0.6, the length after scaling is 1000 × 0.6 = 600, and the width is 800 × 0.6 = 480.

[0071] By analogy, the scaled dimensions of the product image under each scaling factor can be obtained.

[0072] Specifically, please refer to Figure 2 The diagram illustrates the steps for obtaining the contour recognition result of a product image according to an embodiment of the present invention. The process of obtaining the contour recognition result of a product image under different scaling states includes:

[0073] Step S21: Determine the contour recognition lines within each scaled image;

[0074] Step S22: Place the contour recognition lines in each scaled image under the same scale and calculate the overlap of any two contour recognition lines among the several contour recognition lines.

[0075] Step S23: The average value of the overlap of several sets of two contour recognition lines is determined as the quality confidence value of the product image.

[0076] Specifically, the present invention can determine the contour recognition line in each zoomed image based on edge detection and contour extraction technology in computer vision. Edge detection and contour extraction technology in computer vision is a commonly used image processing method in the AI ​​field, which will not be elaborated here.

[0077] Specifically, the overlap of two contour recognition lines can be calculated by normalizing the coordinates of the coordinate points on the contour recognition lines and determining the overlap ratio of the number of overlapping points between the two curves.

[0078] Specifically, when determining the contour recognition line in each scaled image, the outermost closed contour in each scaled image is taken as the representative contour of the image, and the overlap of the two contour recognition lines is determined based on this.

[0079] Specifically, the product image corresponding to the maximum value of the quality confidence characterization value is selected as the quality confidence feature image.

[0080] It is understood that this invention generates product images at different scaling states and extracts contour recognition lines. After conversion to the same scale, the overlap of any two contour lines is calculated and the average value is taken as the quality confidence characterization value. Finally, the image with the largest value is selected as the quality confidence feature image. This is because the high overlap of contour lines at multiple scaling states indicates that the image is less affected by shooting angle, lighting, etc., and can stably reflect the true geometric features of the product, thereby improving the reliability and efficiency of product information processing.

[0081] Specifically, please refer to Figure 3As shown, this is a flowchart illustrating the logic for determining whether to store or delete the supplier's product information. The process of determining whether to store or delete the supplier's product information based on the results of the comparison analysis includes:

[0082] Calculate the intersection-union ratio (IUR) of the quality confidence feature image with each sample image in the existing product database, and determine the maximum IUR value;

[0083] If the maximum value of the cross-union ratio is less than the preset cross-union ratio threshold, it is determined that the product information should be stored.

[0084] If the maximum value of the cross-union ratio is greater than or equal to the preset cross-union ratio threshold, it is determined that the product information should be cleared.

[0085] Specifically, the intersection-union ratio (IUGR) in image processing is an important indicator for measuring the degree of overlap between two regions. It is widely used in fields such as object detection, image segmentation, and region matching. In this invention, the ratio of the intersection area to the union area of ​​the two regions, the quality confidence feature image and the sample image, is calculated. The result ranges from [0, 1]. The closer the IUGR is to 1, the higher the degree of overlap. The method for determining the IUGR in image processing is well known to those skilled in the art, and its calculation formula is also existing technology, so it will not be elaborated here.

[0086] Specifically, the crossover and union ratio (CUP) threshold can be determined based on information filtering requirements. If the scenario requires strict prevention of supplier misuse of product information, the CUP threshold should be set relatively small. Setting the CUP threshold too small can also lead to misjudgment of supplier misuse of product information. Based on this, the range of the CUP threshold can be set to [0.88, 0.92]. This embodiment provides a CUP threshold value of 0.9.

[0087] It is understood that when comparing the quality confidence feature image with the existing sample images in the product database in this embodiment, the images being compared should be the same product placed in the same position and taken from the same angle.

[0088] Understandably, in the practical scenario of supplier product information management in the manufacturing industry, this invention compares the previously screened quality confidence feature images with the existing sample images in the product database, calculates the ratio of the overlapping area of ​​the two contours to the total area, i.e., the intersection-union ratio (IUR), and finds the maximum value to measure the similarity between the new product and the existing products in the database. When the maximum value is lower than a preset threshold, it indicates that the new product is significantly different from the existing samples in the database and belongs to a new specification product. When the calculated maximum IUR of the quality confidence feature image and the sample images in the existing product database reaches or exceeds the preset threshold, it means that the product is highly similar to the existing samples in the database in terms of features, and there is a situation where the supplier's product image information is misusing other suppliers' information. In supply management, there is a phenomenon where the provided product is actually similar to other suppliers' existing products, but attempts to enter the procurement process with new information. Clearing such information can avoid memory occupation caused by information duplication, prevent redundancy in the existing product database from affecting the efficiency of subsequent information processing, and reduce the procurement risks that may be caused by information misuse. It realizes intelligent filtering of supplier product information, balances information integrity and management efficiency, and provides reliable data support for procurement decisions.

[0089] Specifically, the process of determining the first dimension feature index of the product corresponding to the product image in production includes:

[0090] Determine the duration of the timeline between the start time of production and the timeline of production.

[0091] The value obtained by dividing the duration by the total duration of commodity production is determined as the first dimension feature index.

[0092] Specifically, the unit for the time progress node of a product's participation in product production is minutes. For example, if the product production start time node is minute 0 and the product production time progress node is minute 150, that is, 150 minutes after the product production starts, is the node when a certain component is completed or assembled, then the duration of both is calculated as follows: Duration = Product production time progress node - Product production start time node = 150 minutes - 0 minutes = 150 minutes.

[0093] Understandably, in the practical scenarios of production scheduling and procurement coordination in manufacturing, the significance of determining the first-dimensional characteristic index lies in providing an objective time-dimensional basis for prioritizing procurement by quantifying the order of components in the production sequence. Specifically, each link in the production process has a specific time node involved in production. The ratio of the duration of this node from the start of production to the total duration of the entire product's production directly reflects the degree of influence of a certain component's processing or assembly node on the timeline. For example, if a link is on the 10th day after production starts, and the total production cycle is 50 days, its first-dimensional characteristic index is 0.2; if a link is on the 40th day, its first-dimensional characteristic index is 0.8. The smaller the first-dimensional characteristic index, the earlier the product participates in production, which is the basis for the smooth progress of subsequent processes. Delays in the procurement of such products can cause the entire production chain to stagnate prematurely, resulting in huge time and cost losses.

[0094] Specifically, please refer to Figure 4 The diagram illustrates the steps for determining the second-dimensional feature index in an embodiment of the present invention. The process of determining the second-dimensional feature index of the product corresponding to the product image during production includes:

[0095] Step S41: Obtain the product information and extract several performance parameters of the product information;

[0096] Step S42: Compare several performance parameters of the product information with the performance parameters corresponding to historical product information to determine the number of performance parameter groups with differences.

[0097] Step S43: The ratio of the number of performance parameter groups with differences to the total number of performance parameter groups in historical product information is determined as the second dimension feature index.

[0098] Specifically, the performance parameters in this invention may include the operational stability and durability of the assembled product. By comparing these parameters with the corresponding parameters of similar historical products in the existing product library, the groups of parameters that differ can be accurately located, and thus the number of such groups can be determined.

[0099] Understandably, in manufacturing procurement scenarios where product quality and compatibility are controlled, the significance of determining the second-dimensional characteristic indicator lies in quantifying the degree of difference between product performance and historical data. This provides a reference for the level of risk in procurement decisions based on the stability and durability of the assembled product during subsequent use. The smaller the second-dimensional characteristic indicator, the higher the performance overlap between the new product and historical samples, meaning that its technology is mature and highly compatible with the existing production system, resulting in a lower risk of assembly conflicts or quality fluctuations after procurement. Conversely, the larger the second-dimensional characteristic indicator, the more likely the new product has compatibility risks that require careful review. This avoids production risks caused by blind procurement and improves the accuracy of decision-making.

[0100] Specifically, the value obtained by dividing the second dimension feature index by the first dimension feature index is determined as the comprehensive feature index.

[0101] Specifically, the numerical range of the comprehensive characteristic index is (0, 10).

[0102] Understandably, the ratio of the second-dimensional feature indicator to the first-dimensional feature indicator is used as the comprehensive feature indicator. The principle behind this is to determine a comprehensive quantitative indicator by integrating two major factors: the urgency of time and the compatibility of processes. The first-dimensional feature indicator reflects the urgency of the product in the production sequence, while the second-dimensional feature indicator reflects the degree of parameter difference between the product and historical samples. The larger the comprehensive feature indicator, in real-world scenarios, it means that the performance parameters of the product differ significantly from those of similar historical products, which may involve compatibility risks. Furthermore, the product belongs to a critical link in the production process that needs to be put into operation as soon as possible, and has a high time urgency. Therefore, such products need to be prioritized as objects requiring urgent review to ensure that the evaluation is completed within the time window.

[0103] Specifically, the process of determining whether to compress the product image into several data packets and transmit them to the review terminal includes:

[0104] The comprehensive feature index is compared with the preset comprehensive feature index reference value;

[0105] If the comprehensive feature index is greater than the comprehensive feature index reference value, then it is determined that the product image will be compressed into several data packets and transmitted to the review terminal.

[0106] If the comprehensive feature index is less than or equal to the reference value of the comprehensive feature index, it is determined that the product image does not need to be compressed into several data packets for transmission to the review terminal, and the product image can be packaged into a single data packet for transmission.

[0107] Specifically, the reference value of the comprehensive characteristic index can be set based on historical production data or test data. In the implementation of this invention, the reference value of the comprehensive characteristic index can be set according to the statistical analysis results of historical test data. The average value of the second dimension characteristic index divided by the first dimension characteristic index in several tests is pre-tested and recorded. The average value of the second dimension characteristic index divided by the first dimension characteristic index obtained from the test statistics is recorded as the reference value of the comprehensive characteristic index. Here, a value of the reference value of the comprehensive characteristic index is provided, which can be set to 7.

[0108] It is understood that this invention integrates the degree of compatibility difference and time urgency of a product through comprehensive feature indicators. This means that the product has significant compatibility risks in a tight production cycle and requires more timely review and intervention. When the comprehensive feature indicators exceed the set reference values, it indicates that the product information has reached the level that requires priority review. At this time, its image is compressed and transmitted to the review terminal to ensure that high-risk components are evaluated in a timely manner.

[0109] Specifically, the process of compressing the product image into several data packets and transmitting them to the review terminal includes:

[0110] Determine the outline recognition lines within the product image;

[0111] Product images where the contour recognition line intersects with the edge of the product image are compressed into the first data packet, and product images where the contour recognition line does not intersect with the edge of the product image are compressed into the second data packet;

[0112] The first data packet and the second data packet are transmitted to the auditing terminal in sequence.

[0113] Specifically, in the implementation of this invention, the contour recognition line in the product image can be determined based on edge detection and contour extraction technology in computer vision. Edge detection and contour extraction technology in computer vision is a commonly used image processing method in the AI ​​field, which will not be elaborated here.

[0114] Specifically, in the implementation of this invention, the WebP algorithm can be used to compress the product image into the first data packet or the product image into the second data packet. The technical means of compressing and packaging the image for transmission is a common information transmission method used by those skilled in the art, and will not be elaborated here.

[0115] It is understood that this invention distinguishes whether the contour recognition line intersects with the image edge. The global image of the product is a complete presentation of the overall shape of the product. In this case, the contour line of the product is a complete closed curve, and its contour range is completely contained within the image. It will not intersect with the image edge. However, the local image of the product is a close-up of a local area of ​​the product. In this case, the contour line of the product will be cut off by the image edge, resulting in the contour line intersecting with the image edge.

[0116] Specifically, the partial images of the product in the first data packet can show key details of the parts, such as corners, interfaces, and partial fragments of irregular structures. These local features are crucial to the product's assembly compatibility or functional stability and are core information that needs to be prioritized during the review process. The global images of the product in the second data packet present the overall shape of the product, such as its overall layout and the relative positions of its parts. This type of information provides weaker support for judging local details. This invention prioritizes the transmission of the first data packet, allowing the review terminal to quickly verify the product's key performance characteristics. In this invention, comprehensive feature indicators characterize the product's functional compatibility and need to be judged first, while indicators reflecting the overall performance can be verified later. This priority aligns with the transmission order of the data packets. After the first data packet is transmitted, the review terminal can immediately and quickly determine whether the product meets the requirements based on the partial images. If the indicators in the first data packet do not meet the standards, the review can be terminated directly without waiting for the second data packet, reducing unnecessary calculations.

[0117] Specifically, this invention divides a product image into a first data packet containing local key features and a second data packet containing the complete global shape based on whether the contour recognition line intersects with the image edge, and transmits them sequentially. This allows the review terminal to prioritize the acquisition of image information of the local key areas of the product, enabling rapid review of core quality features. At the same time, it can supplement the overall shape verification through the subsequently transmitted global image. While ensuring the integrity of the review, it reduces the occupation of transmission bandwidth and review resources by non-critical information, thereby improving the efficiency and accuracy of product quality review. It is especially suitable for industrial quality inspection scenarios that require both detail accuracy and overall compliance.

[0118] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. An AI-based supplier procurement management method, characterized by, The application comprises the following steps: acquiring the product information sent by the supplier in advance at the information receiving interface, and extracting the product image in the product information; scaling each product image based on the pre-set scaling coefficient, obtaining the contour recognition result of the product image in different scaling states, and screening the quality confidence feature image according to the coincidence of the contour recognition result; wherein the process of scaling each product image based on the pre-set scaling coefficient comprises: pre-setting several scaling coefficients; determining the scaling length-width size of each product image based on each scaling coefficient, so as to determine the scaling state image according to the scaling length-width size; wherein the setting value of the scaling coefficient is in an arithmetic sequence; wherein the process of obtaining the contour recognition result of the product image in different scaling states comprises: determining the contour recognition line in each scaling state image; placing the contour recognition lines in each scaling state image under the same scale, and calculating the coincidence degree of any two contour recognition lines in the contour recognition lines; determining the average value of the coincidence degrees of the several groups of two contour recognition lines as the quality confidence characteristic value of the product image; wherein the product image corresponding to the maximum value of the quality confidence characteristic value is screened as the quality confidence feature image; uploading the quality confidence feature image to the existing product library constructed in advance for comparison and analysis, and determining whether to store or delete the product information of the supplier according to the result of the comparison and analysis; in response to the determination result of storing the product image, determining the first dimension feature index and the second dimension feature index of the product corresponding to the product image in the production, wherein the first dimension feature index is determined according to the time progress node participating in the production of the product, and the second dimension feature index is determined according to the comparison result of the product information and the historical product information; determining the comprehensive feature index of the product based on the first dimension feature index and the second dimension feature index, and determining whether to compress the product image into several data packets and transmit it to the audit terminal based on the comparison result of the comprehensive feature index and the comprehensive feature index reference value; wherein the process of determining the first dimension feature index of the product corresponding to the product image in the production comprises: determining the continuous duration of the time progress node participating in the production of the product from the starting time node of the production of the product; dividing the continuous duration by the total continuous duration of the production of the product to determine the first dimension feature index. 2.The AI-based supplier procurement management method of claim 1, wherein The process of determining whether to store or delete the product information of the supplier according to the comparison and analysis result comprises: calculating the intersection-over-union of the quality confidence feature image and each sample image in the existing product library, and determining the maximum value of the intersection-over-union; if the maximum value of the intersection-over-union is less than the pre-set intersection-over-union threshold value, it is determined to store the product information; if the maximum value of the intersection-over-union is greater than or equal to the pre-set intersection-over-union threshold value, it is determined to delete the product information. 3.The AI-based supplier procurement management method of claim 1, wherein, The process of determining the second dimension feature index of the product corresponding to the product image in the production comprises: extracting several performance parameters of the product information; Compare the performance parameters of the commodity information with the performance parameters corresponding to the historical commodity information, and determine the number of performance parameter groups with differences in performance parameters; Determine the ratio of the number of performance parameter groups with differences to the total number of performance parameters in the historical commodity information as the second dimension feature index. 4.The AI-based supplier procurement management method of claim 1, wherein Determine the value obtained by dividing the second dimension feature index by the first dimension feature index as the comprehensive feature index. 5.The AI-based supplier procurement management method of claim 4, wherein, The process of compressing the commodity image into a plurality of data packets for transmission to the review terminal includes: Compare the comprehensive feature index with the preset comprehensive feature index reference value; If the comprehensive feature index is greater than the comprehensive feature index reference value, it is determined that the commodity image will be compressed into a plurality of data packets for transmission to the review terminal. 6.The AI-based supplier procurement management method of claim 5, wherein, The process of compressing the commodity image into a plurality of data packets for transmission to the review terminal includes: Determine the contour recognition line in the commodity image; Compress the commodity image with intersection between the contour recognition line and the edge of the commodity image into a first data packet, and compress the commodity image without intersection between the contour recognition line and the edge of the commodity image into a second data packet; In turn, transmit the first data packet and the second data packet to the review terminal.

Images