A data tracing method based on supply chain

By integrating image data in the supply chain and calculating the dominant difference coefficients and extracting the dominant texture features, the problem of low detection efficiency of massive image data is solved, efficient image data detection and traceability is achieved, and computing power consumption is reduced.

CN118521327BActive Publication Date: 2024-11-08HUNAN LIAN DONG CLOUD TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410987759.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-23
Publication Date
2024-11-08
Estimated Expiration
2044-07-23

AI Technical Summary

Technical Problem

In the supply chain, the detection efficiency of massive image data is poor and the computing power occupies a high level, resulting in low detection efficiency.

Method used

By obtaining the image data of supply chain products, constructing the relationship between image data and product information, integrating image data and extracting background chromaticity and feature object outlines, calculating dominant difference coefficients for data division, extracting dominant texture features, based on differences Determine the category and select adaptive feature processing methods to improve detection efficiency.

Benefits of technology

It improves the efficiency of detection of massive image data, reduces computing power consumption, and reduces the risk of misjudgment, and is suitable for the detection and traceability of large-scale image data in the supply chain.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118521327B_ABST
    Figure CN118521327B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of image recognition and tracing, and in particular to a data tracing method based on a supply chain. The present invention obtains image data for supply chain products, integrates the image data obtained within a predetermined time, extracts the background chromaticity and characteristic object contours of each of the image data, calculates the dominant difference coefficient based on the morphological difference of the background chromaticity and the characteristic object contour between the image data, divides each of the image data to obtain a plurality of divided data groups, extracts the dominant texture features of each image data in the divided data groups, determines the dominant difference category of the divided data groups based on the difference of the dominant texture features, and adaptively selects a feature processing method for the divided data groups to determine whether the product is abnormal and trace it. The present invention improves the detection and tracing efficiency through the above process, and reduces the computing power consumption when detecting and tracing the massive image data in the supply chain.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of image recognition tracing, and in particular to a data tracing method based on a supply chain. Background Art

[0002] With the development of computer graphics technology, graphic verification technology has been widely used. By extracting relevant features of products in images, it can detect whether products have defects or abnormalities. Especially in the field of supply chain, image verification technology is used to detect supply chain products, trace defective products, and trace relevant product information, which greatly improves efficiency.

[0003] For example, Chinese patent application publication number: CN117236825A, discloses an agricultural product supply chain transportation management system and method, which relates to the field of agricultural products. By obtaining the maturity label of each agricultural product, based on the maturity label, n agricultural product images are collected, n is a positive integer greater than zero, to obtain an agricultural product image set, and the agricultural product image set is used to train and generate a maturity analysis model. Based on different temperature conditions, a first maturity change model is obtained, a real-time agricultural product image set is obtained, and the maturity analysis model is used to obtain the initial maturity of the real-time agricultural products. The maturity and transportation time limit required by the merchant are collected, and the first transportation strategy and the second transportation strategy are generated in combination with the initial maturity of the real-time agricultural products and the first maturity change model, which solves the problem of agricultural products rotting prematurely or maturing too quickly due to inappropriate temperature conditions during transportation.

[0004] However, there are still the following problems in the prior art:

[0005] When performing detection on image data, the corresponding image processing model is often used to extract and analyze relevant features in the image to determine whether there are any abnormalities in the products in the image data. However, when performing detection on the massive amount of image data in the supply chain, the detection efficiency of the image data is not good due to the large amount of data and high computing power usage. Summary of the invention

[0006] To this end, the present invention provides a data tracing method based on a supply chain, so as to overcome the problem in the prior art that when detecting massive image data in the supply chain, the detection efficiency of the image data is poor due to the large amount of data and high computing power consumption.

[0007] To achieve the above object, the present invention provides a data tracing method based on a supply chain, which comprises:

[0008] Step S1, obtaining image data for supply chain products, building an association relationship between the image data and product information and storing it in a database;

[0009] Step S2, integrating the image data acquired within a predetermined time and extracting the background chromaticity and characteristic object contour of each image data, calculating the dominant difference coefficient based on the morphological difference of the background chromaticity and characteristic object contour between the image data, so as to divide each image data to obtain a plurality of divided data groups;

[0010] Step S3, extracting the dominant texture features of each image data in the divided data group, and determining the dominant difference category of the divided data group based on the difference of the dominant texture features;

[0011] Step S4, selecting a feature processing method for the divided data group according to the dominant difference category of the divided data group, the feature processing method includes:

[0012] Extract any image data in the partitioned data group to identify the product category, determine the associated feature extraction requirements, and traverse the partitioned data group based on the feature extraction requirements to extract features of all image data to determine whether the product is abnormal;

[0013] Or, traverse the divided data groups to identify all image data product categories, respectively determine the feature extraction requirements associated with each of the image data and extract features to determine whether the product is abnormal;

[0014] Among them, the correlation between various products and feature extraction requirements is pre-built;

[0015] Step S5, performing a tracing operation on the abnormal image data, wherein the tracing operation includes retrieving product information associated with the image data from the database.

[0016] Furthermore, in step S2, the process of identifying the contour of the characteristic object in the image data includes:

[0017] Calibrate all object contours in the image data;

[0018] Calculate the mean of the degree of fit between each object contour and the contours of the remaining objects respectively;

[0019] The object contour corresponding to the maximum fitting mean is determined as the characteristic object contour.

[0020] Furthermore, in the step S2, in the obtained single divided data group, the dominant difference coefficients between the image data are all smaller than a predetermined dominant difference threshold.

[0021] Furthermore, in step S2, the dominant difference coefficient is calculated according to formula (1):

[0022] ;

[0023] In formula (1), C represents the explicit difference coefficient, P represents the background chromaticity difference between image data, P0 represents the preset background chromaticity standard difference threshold, G represents the fit of the feature object contour between image data, G0 represents the preset fit standard threshold, α represents the chromaticity difference weight coefficient, and β represents the fit difference weight coefficient.

[0024] Furthermore, in step S3, the process of extracting the dominant texture features of the image data includes:

[0025] Extracting images from feature object contours;

[0026] Performing clustering processing on the image to calibrate clusters in the image as texture features;

[0027] The average distribution distance and average area of ​​each grain feature are determined to obtain the dominant grain feature.

[0028] Furthermore, in step S3, the process of determining the dominant difference category of the divided data group based on the difference of dominant texture features includes:

[0029] If the divided data group meets the predetermined condition, the divided data group is determined to be a weak dominant difference category;

[0030] If the divided data group does not meet the predetermined condition, determining the divided data group as a strong dominant difference category;

[0031] The predetermined condition is that the average distribution distance difference ratio of the texture features between any image data is less than a predetermined distribution distance difference ratio difference threshold and the average area difference ratio is less than a predetermined area difference ratio difference threshold.

[0032] Furthermore, in step S3, the feature processing method for the divided data group is selected to include:

[0033] If the divided data group is a weak dominant difference category, then select and extract any image data in the divided data group to identify the product category, determine the associated feature extraction requirements, and traverse the divided data group based on the feature extraction requirements to extract features of all image data to determine whether the product is abnormal;

[0034] If the divided data group is a strongly dominant difference category, the divided data group is selected to be traversed to identify the product categories of all image data, determine the feature extraction requirements associated with each image respectively, and extract features to determine whether the product has an abnormality.

[0035] Furthermore, in the step S4, each type of product is associated with a corresponding feature extraction requirement, and the feature extraction requirement includes a number of features to be extracted from the image data.

[0036] Furthermore, in step S4, the process of extracting features to determine whether the product is abnormal includes:

[0037] The extracted features are compared with a predetermined standard range, and if the features do not fall within the predetermined standard range, it is determined that the product is abnormal;

[0038] The predetermined standard range is a standard range defined in advance for the feature.

[0039] Furthermore, in the step S1, the product information includes a production batch, a production date and a production location for the product corresponding to the image data.

[0040] Compared with the prior art, the present invention obtains image data for supply chain products, integrates the image data obtained within a predetermined time and extracts the background chromaticity and characteristic object contours of each image data, calculates the explicit difference coefficient based on the morphological difference of the background chromaticity and the characteristic object contours between the image data, divides each image data to obtain a number of divided data groups, extracts the explicit texture features of each image data in the divided data groups, determines the explicit difference category of the divided data groups based on the difference of the explicit texture features, and adaptively selects the feature processing method for the divided data groups to determine whether the product is abnormal and trace it. The present invention improves the detection efficiency through the above process and reduces the computing power consumption when detecting and tracing the massive image data in the supply chain.

[0041] In particular, the present invention calculates the explicit difference coefficient and divides the image data groups. In actual situations, due to the particularity of the supply chain, batch products often exist. Similarly, the similarity of a large amount of image data to be detected is relatively high, especially for agricultural products, such as fruits and vegetables. The similarity of the background chromaticity and the contour of the characteristic object between the corresponding image data of a single type of product is relatively high. The purpose of considering the contour of the characteristic object is that there may be products placed in a concentrated manner in the image data. In order to reduce misjudgment, the contour of the characteristic object that can characterize the main characteristics of most products is found, and then the explicit difference coefficient is calculated. The background chromaticity and the contour of the characteristic object are both explicit image features that are easy to extract, and the computing power required for extraction is small. Then, the image data is divided into divided data groups by simple features under the premise of consuming less computing power, which is convenient for subsequent centralized analysis and processing, thereby improving the detection efficiency and reducing the computing power consumption when detecting and tracing the massive image data in the supply chain.

[0042] In particular, the present invention extracts the explicit texture features of each image data in the divided data group, and determines the explicit difference category of the divided data group. In actual situations, especially for fruit and vegetable products, the data representation of the texture features is strong and can represent the product category. The present invention only considers the explicit texture features, that is, the average distribution distance and average area of ​​the texture features. The extraction of the above features consumes little computing power, and has a certain data representation for the texture features. Therefore, based on the difference in the explicit texture features with less computing power consumption, the explicit difference category of the image data in the divided data group can be further determined. Subsequently, different feature processing methods are adopted for the explicit difference category of the divided data group, thereby improving the detection efficiency and reducing the computing power consumption when detecting and tracing the massive image data in the supply chain.

[0043] In particular, the present invention selects a feature processing method according to the category of the divided data group. In actual situations, image recognition models are mostly used for analysis to determine whether there is an abnormality in the product in the image data. First, the category of the product in the image data needs to be identified. When the amount of data is large, the computing power occupancy is high. In addition, the features in the image data are subsequently extracted to detect whether it is qualified. Taking the transportation of various types of fruits in the supply chain as an example, the qualified standards of different fruits are different. For example, some categories of fruits require the chromaticity to be within a predetermined range, and some categories of fruits require the size to be within a predetermined range. Therefore, it is necessary to identify the fruit category, and then extract features from the image data for determination according to the detection requirements required for the category of fruit. Therefore, the present application uses the weakly dominant difference category to divide the data group and identify the product category based on arbitrary image data. Subsequently, the feature extraction requirements are determined to traverse the divided data group to extract features of all image data to determine whether the product is abnormal, thereby improving the detection efficiency and reducing the computing power consumption when detecting and tracing the massive image data in the supply chain. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A schematic diagram of the steps of a data tracing method based on a supply chain according to an embodiment of the invention;

[0045] Figure 2 A logic block diagram for determining the dominant difference categories of a data group according to an embodiment of the present invention;

[0046] Figure 3 A logic block diagram of a method for processing the characteristics of the divided data group selected in an embodiment of the invention;

[0047] Figure 4 The present invention is a logic block diagram for determining whether a product is abnormal. DETAILED DESCRIPTION

[0048] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0049] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0050] See also Figures 1 to 4 As shown, Figure 1 This is a schematic diagram of the steps of a data tracing method based on a supply chain according to an embodiment of the invention. Figure 2 The logical block diagram of determining the dominant difference category of the divided data group according to the embodiment of the present invention is as follows: Figure 3 A logic block diagram of a method for processing the features of the divided data groups selected in an embodiment of the invention, Figure 4 This is a logic block diagram of determining whether a product is abnormal according to an embodiment of the invention. The data tracing method based on the supply chain of the present invention includes:

[0051] Step S1, obtaining image data for supply chain products, building an association relationship between the image data and product information and storing it in a database;

[0052] Step S2, integrating the image data acquired within a predetermined time and extracting the background chromaticity and characteristic object contour of each image data, calculating the dominant difference coefficient based on the morphological difference of the background chromaticity and characteristic object contour between the image data, so as to divide each image data to obtain a plurality of divided data groups;

[0053] Step S3, extracting the dominant texture features of each image data in the divided data group, and determining the dominant difference category of the divided data group based on the difference of the dominant texture features;

[0054] Step S4, selecting a feature processing method for the divided data group according to the dominant difference category of the divided data group, the feature processing method includes:

[0055] Extract any image data in the partitioned data group to identify the product category, determine the associated feature extraction requirements, and traverse the partitioned data group based on the feature extraction requirements to extract features of all image data to determine whether the product is abnormal;

[0056] Or, traverse the divided data groups to identify all image data product categories, respectively determine the feature extraction requirements associated with each of the image data and extract features to determine whether the product is abnormal;

[0057] Among them, the correlation between various products and feature extraction requirements is pre-built;

[0058] Step S5, performing a tracing operation on the abnormal image data, wherein the tracing operation includes retrieving product information associated with the image data from the database.

[0059] Specifically, the purpose of obtaining supply chain product image data is to check whether there are any abnormalities in the products in the image data and to trace them. The image data can be uploaded by the user end, and the image data should have a certain clarity to facilitate subsequent inspection and analysis, which will not be elaborated here.

[0060] Specifically, there is no limitation on the specific method of obtaining the background chromaticity in the image and identifying the contour of the object. The background chromaticity can be the average chromaticity of a single frame of image data, and the edge detection algorithm can be used to identify the contour of the object. This is a prior art and will not be described in detail.

[0061] Specifically, there is no specific limitation on the method of constructing the association relationship between image data and product information. It can be in the form of setting tags, using product information as tags and setting timestamps to associate with image data, which will not be elaborated here.

[0062] Specifically, in step S2, the process of identifying the contour of a characteristic object in the image data includes:

[0063] Calibrate all object contours in the image data;

[0064] Calculate the mean of the degree of fit between each object contour and the contours of the remaining objects respectively;

[0065] The object contour corresponding to the maximum fitting mean is determined as the characteristic object contour.

[0066] Specifically, there is no limitation on the method of obtaining the object contour fitting degree. For example, a shape context matching algorithm may be used to calculate the object contour fitting degree. Of course, other methods may also be used. This is the prior art and will not be described in detail.

[0067] Specifically, there is no limitation on the method for identifying product types in image data. Commonly, a convolutional neural network model with corresponding functions can be pre-trained to identify product types. This is a prior art and will not be described in detail.

[0068] Specifically, in the step S2, in the obtained single divided data group, the dominant difference coefficients between the image data are all smaller than the predetermined dominant difference threshold.

[0069] Specifically, the dominant difference threshold was selected in the interval [1.05, 1.15].

[0070] Specifically, in step S2, the dominant difference coefficient is calculated according to formula (1):

[0071] ;

[0072] In formula (1), C represents the explicit difference coefficient, P represents the background chromaticity difference between image data, P0 represents the preset background chromaticity standard difference threshold, G represents the fit of the feature object contour between image data, G0 represents the preset fit standard threshold, α represents the chromaticity difference weight coefficient, and β represents the fit difference weight coefficient.

[0073] Specifically, the background chromaticity difference threshold is obtained in advance, and several image data belonging to the same type of product are obtained, and the mean background chromaticity difference △P between the image data and the fit ΔG of the characteristic object contour are solved. P0=ΔP×g1 is set, and G0=g2×ΔG is set. g1 represents the color difference offset coefficient, 0.8<g1<0.9, and g2 represents the fit offset coefficient, 0.9<g2<0.95.

[0074] The value of α is 0.4 and the value of β is 0.6.

[0075] Specifically, the present invention calculates the explicit difference coefficient and divides the image data groups. In actual situations, due to the particularity of the supply chain, batch products often exist. Similarly, the similarity of a large amount of image data to be detected is relatively high, especially for agricultural products, such as fruits and vegetables. The similarity of the background chromaticity and the contour of the characteristic object between the corresponding image data of a single type of product is relatively high. The purpose of considering the contour of the characteristic object is that there may be products placed in a concentrated manner in the image data. In order to reduce misjudgment, the contour of the characteristic object that can characterize the main characteristics of most products is found, and then the explicit difference coefficient is calculated. The background chromaticity and the contour of the characteristic object are both explicit image features that are easy to extract, and the computing power required for extraction is small. Then, the image data is divided into divided data groups by simple features under the premise of consuming less computing power, which is convenient for subsequent centralized analysis and processing, thereby improving the detection efficiency and reducing the computing power consumption when detecting and tracing the massive image data in the supply chain.

[0076] Specifically, in step S3, the process of extracting the dominant texture features of the image data includes:

[0077] Extract the image from the contour of the characteristic object;

[0078] Performing clustering processing on the image to calibrate clusters in the image as texture features;

[0079] The average distribution distance and average area of ​​each grain feature are determined to obtain the dominant grain feature.

[0080] Specifically, the present invention does not specifically limit the clustering algorithm for clustering processing. For example, K-means clustering can be used. By clustering the image, clusters in the image, that is, texture features, can be identified. Of course, for the convenience of comparison, clusters with an area smaller than a predetermined threshold need to be eliminated. The predetermined threshold can be selected within the interval [0.5mm, 1mm].

[0081] The average distribution distance of the texture features is the mean of the distances between each cluster and the closest cluster;

[0082] The average area is the average area of ​​each cluster.

[0083] Specifically, in step S3, the process of determining the dominant difference category of the divided data group based on the difference of dominant texture features includes:

[0084] If the divided data group meets the predetermined condition, the divided data group is determined to be a weak dominant difference category;

[0085] If the divided data group does not meet the predetermined condition, determining the divided data group as a strong dominant difference category;

[0086] The predetermined condition is that the average distribution distance difference ratio of the texture features between any image data is less than a predetermined distribution distance difference ratio difference threshold and the average area difference ratio is less than a predetermined area difference ratio difference threshold.

[0087] Specifically, the average distribution distance difference ratio D1 is calculated by formula (2):

[0088] ;

[0089] In formula (2), D represents the absolute value of the average distribution distance difference between image data, and D0 represents the mean value of the average distribution distance between image data.

[0090] The average area difference ratio S1 is calculated by formula (3):

[0091] ;

[0092] In formula (2), S represents the absolute value of the average area difference of the texture features between the image data, and S0 represents the mean of the average area of ​​the texture features between the image data.

[0093] Specifically, the distribution distance difference ratio difference threshold is selected in the interval [0.2, 0.3];

[0094] The area difference ratio threshold is selected in the interval [0.3, 0.5].

[0095] Specifically, the present invention extracts the explicit texture features of each image data in the divided data group, and determines the explicit difference category of the divided data group. In actual situations, especially for fruit and vegetable products, the data representation of the texture features is strong and can represent the product category. The present invention only considers the explicit texture features, that is, the average distribution distance and average area of ​​the texture features. The extraction of the above features consumes little computing power, and has a certain data representation for the texture features. Therefore, based on the difference in the explicit texture features with less computing power consumption, the explicit difference category of the image data in the divided data group can be further determined. Subsequently, different feature processing methods are adopted for the explicit difference category of the divided data group, thereby improving the detection efficiency and reducing the computing power consumption when detecting and tracing the massive image data in the supply chain.

[0096] Specifically, in step S3, the feature processing method for the divided data group is selected to include:

[0097] If the divided data group is a weak dominant difference category, then select and extract any image data in the divided data group to identify the product category, determine the associated feature extraction requirements, and traverse the divided data group based on the feature extraction requirements to extract features of all image data to determine whether the product is abnormal;

[0098] If the divided data group is a strongly dominant difference category, the divided data group is selected to be traversed to identify the product categories of all image data, determine the feature extraction requirements associated with each image respectively, and extract features to determine whether the product has an abnormality.

[0099] The present invention selects a feature processing method according to the category of the divided data group. In actual situations, image recognition models are mostly used for analysis to determine whether there is an abnormality in the product in the image data. First, the category of the product in the image data needs to be identified. When the data volume is large, the computing power occupancy is high. In addition, the features in the image data are subsequently extracted to detect whether it is qualified. Taking the transportation of various types of fruits in the supply chain as an example, the qualified standards of different fruits are different. For example, some categories of fruits require the chromaticity to be within a predetermined range, and some categories of fruits require the size to be within a predetermined range. Therefore, it is necessary to identify the fruit category, and then extract features from the image data for determination according to the detection requirements required for the category of fruit. Therefore, the present application uses the weakly dominant difference category to divide the data group and identify the product category based on arbitrary image data. Subsequently, the feature extraction requirements are determined to traverse the divided data group to extract the features of all image data to determine whether the product is abnormal, thereby improving the detection efficiency and reducing the computing power consumption when detecting and tracing the massive image data in the supply chain.

[0100] Specifically, in step S4, each type of product is associated with a corresponding feature extraction requirement, and the feature extraction requirement includes a number of features to be extracted from the image data.

[0101] Specifically, in step S4, the process of extracting features to determine whether the product is abnormal includes:

[0102] The extracted features are compared with a predetermined standard range, and if the features do not fall within the predetermined standard range, it is determined that the product is abnormal;

[0103] The predetermined standard range is a standard range defined in advance for the feature.

[0104] Specifically, it can be understood that for image data corresponding to different products, the features to be verified are different, and thus the requirements for feature extraction of image data are different;

[0105] Taking the transportation of various types of fruits in the supply chain as an example, different fruits have different qualification standards. For example, some types of fruits need the chromaticity to be within a predetermined range, and some types of fruits need the size to be within a predetermined range, etc. The corresponding features to be extracted in the image data can be the chromaticity of the object's outline, or the area of ​​the object's outline. Therefore, those skilled in the art can set different feature extraction requirements for various types of products according to the inspection requirements, and set the corresponding predetermined standard range for the features to be extracted, which will not be elaborated here.

[0106] Specifically, in step S1, the product information includes the production batch, production date and production location of the product corresponding to the image data. Of course, those skilled in the art can also set other product-related information according to specific needs, which will not be repeated here.

[0107] If the supply chain-based data traceability method of the present invention is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0108] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle 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 fall within the protection scope of the present invention.

Claims

1. A data tracing method based on a supply chain, characterized in that: include: Step S1, obtaining image data for supply chain products, building an association relationship between the image data and product information and storing it in a database; Step S2, integrating the image data acquired within a predetermined time and extracting the background chromaticity and characteristic object contour of each image data, calculating the dominant difference coefficient based on the morphological difference of the background chromaticity and characteristic object contour between the image data, so as to divide each image data to obtain a plurality of divided data groups; Step S3, extracting the dominant texture features of each image data in the divided data group, and determining the dominant difference category of the divided data group based on the difference of the dominant texture features; Step S4, selecting a feature processing method for the divided data group according to the dominant difference category of the divided data group, the feature processing method includes: If the divided data group is a weak dominant difference category, then select and extract any image data in the divided data group to identify the product category, determine the associated feature extraction requirements, and traverse the divided data group based on the feature extraction requirements to extract features of all image data to determine whether the product is abnormal; If the divided data group is a strong dominant difference category, then the divided data group is selected to be traversed to identify the product category of all image data, and the feature extraction requirements associated with each image data are determined respectively and features are extracted to determine whether the product is abnormal; Among them, the correlation between various products and feature extraction requirements is pre-built; Step S5, performing a tracing operation on the abnormal image data, the tracing operation comprising retrieving product information associated with the image data from the database; In step S2, the dominant difference coefficient is calculated according to formula (1): ; In formula (1), C represents the dominant difference coefficient, P represents the background chromaticity difference between image data, P0 represents the preset background chromaticity standard difference threshold, G represents the fit of the feature object contour between image data, G0 represents the preset fit standard threshold, α represents the chromaticity difference weight coefficient, and β represents the fit difference weight coefficient; In step S3, the process of extracting dominant texture features of image data includes: Extracting images from feature object contours; Performing clustering processing on the image to calibrate clusters in the image as texture features; Determine the average distribution distance and average area of ​​each grain feature to obtain the dominant grain feature; In step S3, the process of determining the dominant difference category of the divided data group based on the difference of dominant texture features includes: If the divided data group meets the predetermined condition, the divided data group is determined to be a weak dominant difference category; If the divided data group does not meet the predetermined condition, determining the divided data group as a strong dominant difference category; The predetermined condition is that the average distribution distance difference ratio of the texture features between any image data is less than a predetermined distribution distance difference ratio difference threshold and the average area difference ratio is less than a predetermined area difference ratio difference threshold.

2. The data tracing method based on the supply chain according to claim 1 is characterized in that: In step S2, the process of identifying the contour of a characteristic object in the image data includes: Calibrate all object contours in image data; Calculate the mean of the degree of fit between each object contour and the contours of the remaining objects respectively; The object contour corresponding to the maximum fitting mean is determined as the characteristic object contour.

3. The data tracing method based on supply chain according to claim 1 is characterized in that: In the step S2, in the obtained single divided data group, the dominant difference coefficients between the image data are all smaller than the predetermined dominant difference threshold.

4. The data tracing method based on supply chain according to claim 1 is characterized in that: In the step S4, each type of product is associated with a corresponding feature extraction requirement, and the feature extraction requirement includes a number of features to be extracted from the image data.

5. The data tracing method based on supply chain according to claim 1 is characterized in that: In step S4, the process of extracting features to determine whether the product is abnormal includes: The extracted features are compared with a predetermined standard range, and if the features do not fall within the predetermined standard range, it is determined that the product is abnormal; The predetermined standard range is a standard range defined in advance for the feature.

6. The data tracing method based on supply chain according to claim 1 is characterized in that: In the step S1, the product information includes the production batch, production date and production location of the product corresponding to the image data.