Agricultural product quality safety supervision system

Abstract

The invention, which relates to the technical field of agricultural product quality monitoring, discloses an agricultural product quality safety supervision system comprising a shooting module, a variety identification module, an appearance detection module, a near-infrared detection module, a quality analysis module, a quality standard setting module and a quality evaluation module. The shooting module is used for shooting the agricultural product to obtain an agricultural product image; according to the agricultural product quality safety supervision system, the shooting module, the variety identification module, the appearance detection module, the near-infrared detection module, the quality analysis module, the quality standard setting module and the quality evaluation module are arranged, and the quality information of agricultural products is obtained through analysis by obtaining the near-infrared spectrum data of the agricultural products; whether the appearance of the agricultural product has defects or not is determined by analyzing the appearance image of the agricultural product, so that the quality grade of the agricultural product can be rapidly determined.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural product quality monitoring, and in particular to an agricultural product quality and safety supervision system. Background Art

[0002] Agricultural products refer to primary products derived from planting, forestry, animal husbandry and fishery, that is, plants, animals, microorganisms and their products obtained in agricultural activities, such as rice, sorghum, peanuts, corn, wheat and local specialties of various regions.

[0003] The quality of agricultural products includes many aspects, such as the quality of the agricultural products themselves, pesticide residues, pest conditions, etc. In order to ensure the quality of agricultural products leaving the warehouse, it is necessary to test these aspects of the agricultural products. The general testing method usually requires a large number of tests to test the content of various chemical substances in agricultural products, and it is also necessary to test whether the appearance and color of the agricultural products are qualified. This requires a lot of effort, which increases the difficulty of agricultural product quality supervision. Summary of the invention

[0004] The purpose of the present invention is to provide a system for monitoring the quality and safety of agricultural products to solve the above-mentioned deficiencies in the prior art.

[0005] In order to achieve the above-mentioned purpose, the present invention provides the following technical solutions: an agricultural product quality and safety supervision system, comprising a shooting module, a variety identification module, an appearance detection module, a near infrared detection module, a quality analysis module, a quality standard setting module, and a quality assessment module; The shooting module is used to shoot the agricultural products to obtain images of the agricultural products, wherein the shooting module can use a high-definition camera to capture details of the agricultural products and improve the clarity of the agricultural products images; The variety recognition module is used to identify the types of agricultural products in the agricultural product image based on the target recognition model and obtain corresponding type labels, wherein a pre-trained target recognition model can be selected to identify the types of agricultural products in the agricultural product image and output the type labels corresponding to the identified types; The appearance detection module is used to mark the defect position of the agricultural product in the agricultural product image, obtain a defect image corresponding to the defect position, and set a plurality of defect labels to associate the defect labels with the corresponding defect images; The appearance detection module is also used to train a deep learning model based on the agricultural product image, defect image and defect label to obtain an appearance defect recognition model for identifying and outputting defect images and defect labels corresponding to the defect images based on the input agricultural product images; The near-infrared detection module is used to detect the agricultural products based on the near-infrared spectrum detection technology to obtain the near-infrared spectrum data reflected or transmitted by the agricultural products; wherein, the near-infrared spectrum detection technology is generally divided into two measurement modes: reflection spectrum and transmission spectrum; when the reflection mode is used to detect the quality of fruits, the light source and the optical fiber are placed on the same side to measure the reflection of the light by the detection object. When the transmission mode is used to detect the quality of fruits, the detection optical fiber is placed opposite the light source to collect the transmitted light after penetrating the fruit sample; the collected fruit near-infrared spectrum data and the corresponding physical and chemical indicators are used as the original data, and the qualitative analysis or quantitative analysis is performed according to the needs of the detection indicators by means of the chemometric method, thereby realizing the non-destructive detection of the fruit's soluble solid content (SSC), acidity (pH value), moisture content (Moisture) and other indicators.

[0006] The quality analysis module is used to obtain quality information of agricultural products, wherein the quality information includes but is not limited to surface hardness, internal moisture content, sugar content, acidity and vitamin content. Based on the quality information and near-infrared spectral data of the agricultural products, a first supervised learning algorithm is trained to obtain a quality analysis algorithm that outputs corresponding quality information based on the input near-infrared spectral data; wherein, when training the algorithm, the near-infrared spectral data needs to be preprocessed, such as baseline correction, spectral smoothing, spectral filtering and other operations to improve the quality and accuracy of the data; the present invention does not limit the specific algorithm or model. In this embodiment, linear or nonlinear algorithms such as partial least squares regression (PLS), multivariate linear regression (MLR), principal component regression (PCR), support vector regression (SVR), artificial neural network (ANN), cluster analysis (CA) and partial least squares discriminant analysis (PLS-DA) are generally used.

[0007] The quality standard setting module is used to set a first quality grade, and correspond and associate quality information with the first quality grade according to different types of agricultural products; The quality assessment module is used to obtain a first quality grade corresponding to the type and quality information of the agricultural product based on the quality information output by the quality analysis algorithm and the type of the agricultural product corresponding to the quality information.

[0008] Furthermore, the appearance detection module is also used to obtain an agricultural product image of a normal agricultural product, obtain the normal agricultural product image, extract the color of the agricultural product in the normal agricultural product image, and obtain the normal agricultural product color, including the following steps: Obtaining agricultural product images of normal agricultural products; Classifying the agricultural product images according to the types of agricultural products to obtain multiple agricultural product image sets; Classifying the agricultural product image set according to the first quality grade of the agricultural product, and dividing each agricultural product image set into a plurality of agricultural product image subsets; The colors of the agricultural products of the normal agricultural product images in each agricultural product image subset are extracted respectively to obtain the normal agricultural product colors of the agricultural product image subset.

[0009] Furthermore, because each agricultural product of the first quality grade may contain multiple colors, that is, each of the agricultural product image subsets may contain multiple normal agricultural product colors, the normal product color of each agricultural product image subset may be represented in the form of a color interval, including the following steps: Get the Lab value of the normal agricultural product color of each agricultural product image subset; Screen out the outliers in the Lab values ​​of each subset of agricultural product images; Based on the Lab values ​​of the outliers screened out, the Lab value interval of the agricultural product image subset is obtained.

[0010] Further, the appearance detection module is also used to set a color anomaly judgment label, and based on the agricultural product image, the type label, the first quality grade, the normal agricultural product color and the color anomaly judgment label of the agricultural product, train a second supervised learning algorithm to obtain a color anomaly judgment algorithm for outputting a corresponding color anomaly judgment label according to the input agricultural product image, the type label, and the first quality grade, including the following steps: Setting a color abnormality judgment label, wherein the color abnormality judgment label includes a normal color label and a color abnormality label; Obtaining a first quality grade and a Lab value interval of the agricultural product corresponding to the agricultural product image, and obtaining the Lab value of the agricultural product image; Determining whether the Lab value of the agricultural product image is within the Lab value range; The judgment result "yes" is associated with a normal color label, and the judgment result "no" is associated with an abnormal color label; The agricultural product image, the type label, and the first quality grade are used as sample inputs, and the corresponding color abnormality judgment label is used as sample output, wherein the sample input and the sample output constitute a training sample; Based on the training samples, a second supervised learning algorithm is trained to obtain a color anomaly judgment algorithm, wherein during training, the training samples are divided into a training set, a validation set, and a test set at a ratio of 7:2:1, and only the training set is used to adjust the parameters of the model during the training phase; then the validation set is used to verify the performance of the model to help adjust the hyperparameters and prevent overfitting. If the training is completed and the optimal hyperparameters have been determined, the test set is used to evaluate the accuracy, recall rate, F1-score and other indicators of the model, and analyze the performance of the model on unseen data, wherein the data of the test set should remain unused throughout the training and validation process to ensure the objectivity and generalization ability of the evaluation results.

[0011] Furthermore, the quality assessment module is also used for a color abnormality judgment algorithm to judge whether the color of the agricultural product corresponds to the quality, including the following steps: Obtaining the Lab value of the agricultural product image, the type label of the corresponding agricultural product, and the first quality grade, and inputting them into the color anomaly judgment algorithm; Determine whether the color anomaly algorithm outputs a color normal label; If not, a color abnormality reminder is issued. The technology of this embodiment can prevent some merchants from applying wax, fumigating sulfur, or soaking agricultural products in medicine to change the color of agricultural products and pass them off as good ones, thereby improving the appearance of agricultural products.

[0012] Furthermore, the quality standard setting module is also used to set a second quality grade, and correspond the second quality grade to the agricultural product according to the defect label corresponding to the agricultural product and the first quality grade.

[0013] Furthermore, the quality assessment module is also used to obtain a second quality grade corresponding to the agricultural product with the first quality grade and the defect label based on the first quality grade and the defect label of the agricultural product.

[0014] Furthermore, the variety recognition module is also used to train a target recognition model based on the agricultural product image, including the following steps: Select a set number of agricultural product images and associate the agricultural product images with corresponding category labels; The agricultural product image is used as the input of the target recognition model, the type label associated with the agricultural product image is used as the output of the target recognition model, the target recognition model is trained, and the target recognition model parameters are determined or updated so that the trained target recognition model is more suitable for the recognition of agricultural products of the agricultural product type corresponding to the agricultural product image.

[0015] 1. Compared with the prior art, the agricultural product quality and safety supervision system provided by the present invention, by setting a shooting module, a variety identification module, an appearance detection module, a near-infrared detection module, a quality analysis module, a quality standard setting module, and a quality assessment module, obtains near-infrared spectrum data of agricultural products, analyzes and obtains quality information of agricultural products, such as surface hardness, internal moisture content, sugar content, acidity and vitamin content, etc., and then determines whether there are defects in the appearance of agricultural products by analyzing the appearance images of agricultural products, thereby quickly determining the quality grade of agricultural products.

[0016] 2. Compared with the prior art, the agricultural product quality and safety supervision system provided by the present invention also collects color information of agricultural products of various qualities, determines the normal color range of agricultural products of various qualities, compares the color information of the agricultural product images actually taken with the normal color range corresponding to the quality grade measured by the corresponding agricultural products, and judges whether the color of the agricultural product corresponds to the quality grade, thereby preventing some merchants from fumigating agricultural products with sulfur, applying wax, or soaking them in medicine to change the color and pass them off as good ones. There is no need to monitor the presence of chemical substances on agricultural products through experiments, thereby reducing the difficulty of detection and improving the detection efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0018] Figure 1 A system structure diagram provided for an embodiment of the present invention; Figure 2 A diagram of steps for obtaining normal agricultural product colors provided by an embodiment of the present invention; Figure 3 A diagram of the steps for training a color anomaly judgment algorithm provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0020] In the description of the present invention, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined. In addition, the terms "installed", "connected" and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0021] Example embodiments will be described more fully below with reference to the accompanying drawings, but the example embodiments may be embodied in different forms and should not be construed as limited to the embodiments set forth herein. On the contrary, the purpose of providing these embodiments is to make the present disclosure thorough and complete and to enable those skilled in the art to fully understand the scope of the present disclosure.

[0022] In the absence of conflict, the various embodiments of the present disclosure and the various features therein may be combined with each other.

[0023] As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0024] The terms used herein are only used to describe specific embodiments and are not intended to limit the present disclosure. As used herein, the singular forms "a", "an" and "the" are also intended to include the plural forms, unless the context clearly indicates otherwise. It will also be understood that when the terms "comprising" and / or "made of" are used in this specification, the presence of the features, wholes, steps, operations, elements and / or components is specified, but the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or groups thereof is not excluded.

[0025] See also Figure 1-Figure 3 , an agricultural product quality and safety supervision system, including a shooting module, a variety identification module, an appearance detection module, a near infrared detection module, a quality analysis module, a quality standard setting module, and a quality assessment module; The shooting module is used to shoot agricultural products to obtain images of agricultural products. The shooting module can use a high-definition camera to capture details of the agricultural products and improve the clarity of the agricultural product images. The variety recognition module is used to identify the types of agricultural products in the agricultural product image based on the target recognition model and obtain the corresponding type labels, wherein a pre-trained target recognition model can be selected to identify the types of agricultural products in the agricultural product image and output the type labels corresponding to the identified types; the variety recognition module is also used to train the target recognition model based on the agricultural product image, including the following steps: (1) Select a set number of agricultural product images and associate the agricultural product images with corresponding category labels; (2) Using the agricultural product image as the input of the target recognition model and the category label associated with the agricultural product image as the output of the target recognition model, the target recognition model is trained, and the target recognition model parameters are determined or updated so that the trained target recognition model is more suitable for the recognition of the agricultural product category corresponding to the agricultural product image.

[0026] The appearance inspection module is used to mark the defective position of the agricultural product in the agricultural product image, obtain the defective image corresponding to the defective position, and set a variety of defect labels to associate the defect labels with the corresponding defective images; The appearance inspection module is also used to train a deep learning model based on the agricultural product image, the defect image and the defect label to obtain an appearance defect recognition model for identifying and outputting the defect image and the defect label corresponding to the defect image based on the input agricultural product image; The near-infrared detection module is used to detect agricultural products based on near-infrared spectroscopy detection technology to obtain near-infrared spectral data reflected or transmitted by agricultural products. Among them, near-infrared spectroscopy detection technology is usually divided into two measurement modes: reflection spectrum and transmission spectrum. When using the reflection mode to detect fruit quality, the light source and the optical fiber are placed on the same side to measure the reflection of the light by the detection object. When using the transmission mode to detect fruit quality, the detection optical fiber is placed opposite the light source to collect the transmitted light after penetrating the fruit sample. The collected fruit near-infrared spectral data and corresponding physical and chemical indicators are used as raw data. With the help of chemometric methods, qualitative or quantitative analysis is performed according to the needs of the detection indicators, thereby realizing non-destructive detection of fruit soluble solid content (SSC), acidity (pH value), moisture content and other indicators.

[0027] The quality analysis module is used to obtain quality information of agricultural products, wherein the quality information includes but is not limited to surface hardness, internal moisture content, sugar, acidity and vitamin content. The quality information for training the first supervised learning algorithm can be obtained through experimental testing. Based on the quality information of agricultural products and near-infrared spectral data, the first supervised learning algorithm is trained to obtain a quality analysis algorithm that outputs corresponding quality information based on the input near-infrared spectral data; wherein, when training the algorithm, the near-infrared spectral data needs to be preprocessed, such as baseline correction, spectral smoothing, spectral filtering and other operations to improve the quality and accuracy of the data; the present invention does not limit the specific algorithm or model. In this embodiment, linear or nonlinear algorithms such as partial least squares regression (PLS), multivariate linear regression (MLR), principal component regression (PCR), support vector regression (SVR), artificial neural network (ANN), cluster analysis (CA) and partial least squares discriminant analysis (PLS-DA) are generally used.

[0028] The quality standard setting module is used to set the first quality grade, and correspond and associate the quality information with the first quality grade according to different types of agricultural products; The quality assessment module is used to obtain a first quality grade corresponding to the type of agricultural product and the quality information based on the quality information output by the quality analysis algorithm and the type of agricultural product corresponding to the quality information.

[0029] The appearance detection module is also used to obtain an agricultural product image of a normal agricultural product, obtain the normal agricultural product image, extract the color of the agricultural product in the normal agricultural product image, and obtain the normal agricultural product color, including the following steps: (1) Obtaining agricultural product images of normal agricultural products; (2) Classifying agricultural product images according to the types of agricultural products to obtain multiple agricultural product image sets; (3) classifying the agricultural product image sets according to the first quality grade of the agricultural products, and dividing each agricultural product image set into a plurality of agricultural product image subsets; (4) The color of the agricultural product of the normal agricultural product image in each agricultural product image subset is extracted respectively to obtain the normal agricultural product color of the agricultural product image subset.

[0030] Because each agricultural product of the first quality level may contain multiple colors, that is, each agricultural product image subset may contain multiple normal agricultural product colors, the normal product color of each agricultural product image subset can be represented in the form of a color interval, including the following steps: (1) Obtain the Lab value of the normal agricultural product color for each agricultural product image subset; (2) Screen out outliers in the Lab values ​​of each subset of agricultural product images; (3) Based on the Lab values ​​of the outliers, the Lab value interval of the agricultural product image subset is obtained.

[0031] The appearance detection module is also used to set a color anomaly judgment label, and based on the agricultural product image, the type label, the first quality grade, the normal agricultural product color and the color anomaly judgment label of the agricultural product, train a second supervised learning algorithm to obtain a color anomaly judgment algorithm for outputting a corresponding color anomaly judgment label according to the input agricultural product image, the type label, and the first quality grade, including the following steps: (1) Setting color anomaly judgment labels, which include normal color labels and abnormal color labels; (2) Obtaining the first quality grade and Lab value range of the agricultural product corresponding to the agricultural product image, and obtaining the Lab value of the agricultural product image; (3) Determine whether the Lab value of the agricultural product image is within the Lab value range; (4) Associating the judgment result "yes" with a normal color label, and associating the judgment result "no" with an abnormal color label; (5) The agricultural product image, category label, and first quality grade are used as sample inputs, and the corresponding color anomaly judgment label is used as sample output. The sample input and sample output constitute the training sample; (6) Based on the training samples, the second supervised learning algorithm is trained to obtain the color anomaly judgment algorithm. During the training, the training samples are divided into training set, validation set and test set in a ratio of 7:2:1. During the training phase, only the training set is used to adjust the parameters of the model; then the validation set is used to verify the performance of the model to help adjust the hyperparameters and prevent overfitting. If the training is completed and the optimal hyperparameters have been determined, the test set is used to evaluate the accuracy, recall rate, F1-score and other indicators of the model, and analyze the performance of the model on unseen data. The data of the test set should remain unused throughout the training and validation process to ensure the objectivity and generalization ability of the evaluation results.

[0032] The quality assessment module is also used in the color anomaly judgment algorithm to determine whether the color of agricultural products corresponds to the quality, including the following steps: (1) Obtain the Lab value of the agricultural product image, the corresponding agricultural product type label, and the first quality grade, and input them into the color anomaly judgment algorithm; (2) Determine whether the color anomaly algorithm outputs a normal color label; (3) If not, a color abnormality reminder is issued. The technology of this embodiment can prevent some merchants from applying wax, fumigating sulfur, or soaking agricultural products in liquid medicine to change the color of agricultural products and pass them off as good ones, thereby improving the appearance of agricultural products.

[0033] The quality standard setting module is also used to set a second quality grade, and to correspond the second quality grade to the agricultural product according to the defect label corresponding to the agricultural product and the first quality grade.

[0034] The quality assessment module is also used to obtain a second quality grade corresponding to the agricultural product with the first quality grade and the defect label based on the first quality grade and the defect label of the agricultural product.

[0035] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A system for monitoring the quality and safety of agricultural products, characterized by: It includes shooting module, variety identification module, appearance detection module, near infrared detection module, quality analysis module, quality standard setting module and quality assessment module; The shooting module is used to shoot the agricultural products to obtain images of the agricultural products; The variety recognition module is used to identify the types of agricultural products in the agricultural product image based on the target recognition model and obtain corresponding type labels; The appearance detection module is used to mark the defect position of the agricultural product in the agricultural product image, obtain a defect image corresponding to the defect position, and set a plurality of defect labels to associate the defect labels with the corresponding defect images; The appearance detection module is also used to train a deep learning model based on the agricultural product image, defect image and defect label to obtain an appearance defect recognition model for identifying and outputting defect images and defect labels corresponding to the defect images based on the input agricultural product images; The near infrared detection module is used to detect the agricultural products based on near infrared spectrum detection technology to obtain near infrared spectrum data reflected or transmitted by the agricultural products; The quality analysis module is used to obtain quality information of agricultural products, and based on the quality information of the agricultural products and near infrared spectrum data, train a first supervised learning algorithm to obtain a quality analysis algorithm that outputs corresponding quality information according to the input near infrared spectrum data; The quality standard setting module is used to set a first quality grade, and correspond and associate quality information with the first quality grade according to different types of agricultural products; The quality assessment module is used to obtain a first quality grade corresponding to the type and quality information of the agricultural product based on the quality information output by the quality analysis algorithm and the type of the agricultural product corresponding to the quality information.

2. The agricultural product quality and safety supervision system according to claim 1, characterized in that: The appearance detection module is also used to obtain an agricultural product image of a normal agricultural product to obtain the normal agricultural product image, extract the color of the agricultural product in the normal agricultural product image to obtain the normal agricultural product color, including the following steps: Obtaining agricultural product images of normal agricultural products; Classifying the agricultural product images according to the types of agricultural products to obtain multiple agricultural product image sets; Classifying the agricultural product image set according to the first quality grade of the agricultural product, and dividing each agricultural product image set into a plurality of agricultural product image subsets; The colors of the agricultural products of the normal agricultural product images in each agricultural product image subset are extracted respectively to obtain the normal agricultural product colors of the agricultural product image subset.

3. The agricultural product quality and safety supervision system according to claim 2, characterized in that: Each of the agricultural product image subsets includes a plurality of normal agricultural product colors, and the normal product color of each agricultural product image subset is represented in the form of a color interval, including the following steps: Get the Lab value of the normal agricultural product color of each agricultural product image subset; Screen out the outliers in the Lab values ​​of each subset of agricultural product images; Based on the Lab values ​​of the outliers screened out, the Lab value interval of the agricultural product image subset is obtained.

4. The agricultural product quality and safety supervision system according to claim 3 is characterized by: The appearance detection module is also used to set a color anomaly judgment label, and based on the agricultural product image, the type label, the first quality grade, the normal agricultural product color and the color anomaly judgment label of the agricultural product, train a second supervised learning algorithm to obtain a color anomaly judgment algorithm for outputting a corresponding color anomaly judgment label according to the input agricultural product image, the type label, and the first quality grade, including the following steps: Setting a color abnormality judgment label, wherein the color abnormality judgment label includes a normal color label and a color abnormality label; Obtaining a first quality grade and a Lab value interval of the agricultural product corresponding to the agricultural product image, and obtaining the Lab value of the agricultural product image; Determining whether the Lab value of the agricultural product image is within the Lab value range; The judgment result "yes" is associated with a normal color label, and the judgment result "no" is associated with an abnormal color label; The agricultural product image, the type label, and the first quality grade are used as sample inputs, and the corresponding color abnormality judgment label is used as sample output, wherein the sample input and the sample output constitute a training sample; A second supervised learning algorithm is trained based on the training samples to obtain a color anomaly judgment algorithm.

5. The agricultural product quality and safety supervision system according to claim 4, characterized in that: The quality assessment module is also used for the color abnormality judgment algorithm to judge whether the color of the agricultural product corresponds to the quality, including the following steps: Obtaining the Lab value of the agricultural product image, the type label of the corresponding agricultural product, and the first quality grade, and inputting them into the color anomaly judgment algorithm; Determine whether the color anomaly algorithm outputs a color normal label; If not, a color abnormality reminder is issued.

6. The agricultural product quality and safety supervision system according to claim 1, characterized in that: The quality standard setting module is also used to set a second quality grade, and to correspond the second quality grade to the agricultural product according to the defect label corresponding to the agricultural product and the first quality grade.

7. The agricultural product quality and safety supervision system according to claim 6, characterized in that: The quality assessment module is also used to obtain a second quality grade corresponding to the agricultural product with the first quality grade and the defect label based on the first quality grade and the defect label of the agricultural product.

8. The agricultural product quality and safety supervision system according to claim 1, characterized in that: The variety recognition module is also used to train a target recognition model based on the agricultural product image, including the following steps: Select a set number of agricultural product images and associate the agricultural product images with corresponding category labels; The agricultural product image is used as the input of the target recognition model, the category label associated with the agricultural product image is used as the output of the target recognition model, the target recognition model is trained, and the parameters of the target recognition model are determined or updated.

Images