Green vegetable quality safety traceability management system

Abstract

The invention discloses a green vegetable quality safety traceability management system, and particularly relates to the field of agricultural Internet of Things, the system comprises a production information acquisition module, a harvesting information acquisition module, a detection and packaging management module and a quality and safety management module; the production information acquisition module is used for recording and managing key data in a production process from a green vegetable production source; the harvesting information acquisition module records and manages data of the green vegetables in the receiving process by using the RFID technology; the detection and packaging management module detects the quality of the green vegetables through a process and packages qualified products; and the quality and safety management module is used for formulating, implementing and monitoring production standards in real time. According to the invention, five links of green vegetable production, harvesting, detection and packaging can be rapidly identified and recorded, and an efficient and transparent solution is provided for green vegetable production.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural Internet of Things, and more specifically, to a green vegetable quality and safety traceability management system. Background Art

[0002] As an important part of our daily diet, the quality and safety of green vegetables are directly related to the health of consumers. Therefore, the agricultural product quality and safety traceability system has emerged.

[0003] The existing open literature 1 (Design and Implementation of Quality and Safety Traceability System for Fresh Vegetable Products, 2017) proposed a quality and safety traceability system for fresh vegetable products, such as Figure 2 As shown, the fresh vegetable product quality and safety traceability system uses the Internet of Things, QR codes and database technology to achieve full-process information traceability from planting, purchasing, testing to processing. The fresh vegetable product quality and safety traceability system consists of hardware equipment (greenhouse sensors, cameras, etc.), data transmission modules (GPRS network) and software platforms (greenhouse management, expert decision-making), and is divided into six core functional modules: base distribution, greenhouse status, traceability management, farming management, expert online and system settings. Through QR code technology, consumers can scan the code to query product batches, planting records, processing information, etc., to enhance market transparency; enterprises and farmers cooperate to standardize production processes, combine real-time environmental monitoring and expert decision support, improve vegetable quality and reduce management costs. However, the QR code in the fresh vegetable product quality and safety traceability system has a limited amount of information storage, making it difficult to record detailed information on the entire vegetable production process. It can only provide batch-level traceability and cannot accurately locate specific production units.

[0004] The existing public document 2 (Development and Application of Vegetable Base Management System Based on Traceability Technology, 2017) proposes a vegetable base management system based on traceability technology, such as Figure 3 As shown, the vegetable base management system based on traceability technology uses "plot + crop + planting date" as the traceability unit, and designs an internal management traceability code (26 bits) and an external query traceability code (20 bits) to achieve full-chain information traceability from planting to consumption. The functions on the enterprise side include base management, harvest management, packaging management, etc., and support the digitization of agricultural records, agricultural materials management, environmental testing and other processes; the consumer side can query the origin, planting environment, responsible person, test report and other information of vegetables by scanning the QR code. However, the vegetable base management system based on traceability technology requires scanning one by one, which is inefficient and does not use rapid identification of large quantities of vegetable products.

[0005] Therefore, there is an urgent need for a traceability management system that has a large storage capacity, is reusable, and can quickly identify and record the production, harvesting, testing, and packaging links of green vegetables. Summary of the Invention

[0006] In order to overcome the above-mentioned defects of the prior art, the present invention provides a green vegetable quality and safety traceability management system, which records greenhouse environmental data and production operation information through RFID tags, and links production history with harvest details, so as to quickly trace problems to specific production links. A TPF-CNN model is proposed to quickly identify pests on green vegetables, reducing errors and time consumption in manual recording, so as to solve the problems raised in the above-mentioned background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A green vegetable quality and safety traceability management system includes a production information collection module, a harvest information collection module, a testing and packaging management module, and a quality and safety management module. The production information collection module is used to record and manage key data in the production process starting from the source of green vegetable production. The harvest information collection module uses RFID technology to record and manage data during the harvest of green vegetables. The testing and packaging management module detects the quality of green vegetables and packages qualified products. The quality and safety management module formulates, implements, and monitors production standards in real time.

[0009] In the inspection and packaging management module, inspectors use RFID readers to scan RFID tags on harvest trays to obtain production and harvest history information for the vegetables. The module then performs quality inspections on the vegetables, checking for pests and pesticide residues. Pests are detected using the TPF-CNN model, and pesticide residue detectors are used to measure pesticide levels. Finally, inspectors package vegetables with pesticide residues below 0.05 mg / kg and those free of pests. Each batch of vegetables is affixed with an RFID tag containing traceability information, which stores data from production, harvesting, and inspection. The TPF-CNN model detects pests using the following specific procedures:

[0010] Step S1, photographing the vegetables with a camera to obtain vegetable image data;

[0011] Step S2, preprocessing the captured original image to improve image quality and reduce interference, wherein the preprocessing includes cropping, scaling, and contrast adjustment;

[0012] In step S3, the TPF-CNN model performs feature extraction and pest detection on the preprocessed image.

[0013] As a further embodiment of the present invention, the TPF-CNN model is a 28-layer convolutional neural network combined with a transition probability function for detecting pests. The transition probability function first converts the continuous features of the input image into discrete features. Discretization reduces the complexity of the features, allowing the convolutional neural network to more efficiently detect and classify pests. The specific steps are as follows:

[0014] Step W1, extracting the continuous features to be processed from the image;

[0015] Step W2: Divide the value range of the continuous feature into several discrete states. For color values, the grayscale range of 0-255 is divided into 10 intervals, each representing a state. For texture features, different states are defined based on the range of gradient intensity.

[0016] Step W3: For adjacent pixels or regions in the image, calculate the transition probability between their characteristic states. If the grayscale value of a pixel belongs to state A and the grayscale value of its adjacent pixel to the right belongs to state B, then record a transition from A to B. Count the state transition frequencies of all adjacent pixel pairs in the entire image and normalize them to probability values.

[0017] Step W4: Use the calculated transition probability as a new feature representation.

[0018] As a further solution of the present invention, the first 20 layers of the TPF-CNN model are composed of 10 groups of "convolutional layers and pooling layers", each group containing a convolutional layer and a pooling layer; the convolution kernel size of each convolutional layer is 3x3, the step size is 1, and the size of each pooling layer is 2x2, the step size is 2; the middle 5 layers of the TPF-CNN model are all convolutional layers with a convolution kernel size of 3x3 and a step size of 1; the last 3 layers of the TPF-CNN model are composed of 2 fully connected layers and 1 output layer, the input of the first fully connected layer is 512 one-dimensional vectors, and the output size is 4096 one-dimensional vectors; the input of the second fully connected layer is 4096 one-dimensional vectors, and the output is 4096 one-dimensional vectors; the input of the output layer is 4096 one-dimensional vectors, and the output is 6 one-dimensional vectors, of which 5 one-dimensional vectors correspond to 5 pest category labels, and the last one-dimensional vector corresponds to the "no pest" label.

[0019] As a further aspect of the present invention, a production information collection module is used to record and manage key data from the very beginning of vegetable production. This module includes the following specific features: The production information collection module utilizes radio frequency identification (RFID) technology to automate the collection and management of production data, enhancing the transparency and traceability of the production process. Farmers first use an RFID reader to scan the greenhouse's RFID tag to obtain a unique greenhouse ID. Next, farmers need to record various data from the production process, including environmental data and production operation records. Environmental data includes soil fertility (such as nitrogen, phosphorus, and potassium content), humidity (soil moisture percentage), and temperature (real-time greenhouse temperature). This environmental data directly impacts the quality of vegetable growth and is obtained through sensors or manual measurement. Production operation records include fertilization records, pest and disease control measures, and irrigation records. Farmers need to enter detailed information such as fertilizer type, amount, and application time, including the specific amount and date. Farmers also need to register the name, application time, and dosage of pesticides to monitor compliance with chemical safety standards. Finally, farmers record irrigation information, including irrigation time and water usage.

[0020] After data entry is complete, the environmental and production operation records are not only stored in the system but also written to the greenhouse's RFID tag. These RFID tags, which use a high-frequency 13.56MHz passive tag, record the greenhouse number and key production data. Simultaneously, all data is uploaded to a MySQL database for centralized management and real-time query.

[0021] As a further solution of the present invention, the harvest information collection module uses RFID technology to record and manage data of the vegetables during the harvest process, including the following specific contents: The harvest information collection module includes three steps for harvesting vegetables, including the following contents:

[0022] In step Y1, the farmer uses an RFID reader to scan the RFID tag attached to the greenhouse. This RFID tag can be read without direct line of sight, speeding up the scanning process. The RFID tag contains a detailed production history of the vegetables during the growing phase. This history includes the environmental conditions (temperature, humidity, and soil quality) used to grow the vegetables, as well as the fertilizers, pesticides, and irrigation used during the growing period.

[0023] In step Y2, farmers scan the RFID tags on the harvest trays used to collect the vegetables during harvest. Each harvest tray is equipped with a unique RFID tag that serves as an identifier for that particular batch of vegetables. After scanning the tray tag, farmers enter additional harvest details directly into the system. These details include the exact time and date of harvest, the specific picking location within the greenhouse or on the farm, and the quantity of vegetables harvested.

[0024] In step S3, the production history is linked to the harvest information to create a record that links the entire life cycle of the greens. This combined data is then stored in two locations: the RFID tag on the harvest tray and a central MySQL database.

[0025] Linking production history with harvest information allows any issues detected at a later stage to be traced not only to harvest conditions but also to specific production practices. By automatically collecting data through RFID scanning and direct input, the harvest information collection module reduces the need for manual record-keeping, which is time-consuming and prone to errors.

[0026] As a further aspect of the present invention, the quality and safety management module includes the following specific contents by formulating, implementing, and monitoring production standards in real time: The quality and safety management module includes three steps, combining standardized management with real-time monitoring to ensure the high quality and safety of green vegetable products. The first step is for technical personnel to formulate detailed production, harvesting, and testing standards based on the latest national and industry standards, including specific requirements for the greenhouse environment; at the same time, they clarify the limits on the use of pesticides and fertilizers; in addition, they include the selection of harvest timing and specific operating procedures for packaging. The second step is for production managers to strictly review the standards formulated by technical personnel, focusing on evaluating their scientificity, practicality, and operability. The third step is for the system to monitor the entire production process in real time. With the help of sensors and RFID technology, the system can automatically collect greenhouse environmental data, pesticide usage records, and harvest information, and compare them with preset standards in real time. If an abnormality is detected, the system will immediately issue an alarm to notify management and operators to take timely adjustment measures.

[0027] The technical effects and advantages of the green vegetable quality and safety traceability management system of the present invention: The present invention records greenhouse environmental data (such as temperature, humidity, soil quality) and production operation information (such as fertilizer and pesticide use) through RFID tags, and links production history with harvest details to ensure data transparency and traceability, overcoming the limitations of existing systems that cannot accurately locate production units and are inefficient. In addition, the system integrates the TPF-CNN model for pest detection, further ensuring the health of vegetables and reducing errors and time-consuming manual records. Its technical effects and advantages include: automation and efficiency, automatic data collection through RFID, suitable for large-scale production; accurate traceability, which can quickly trace problems to specific production links; quality control, the CNN model improves the accuracy of disease and pest detection; real-time monitoring and transparency, supporting real-time management of the production process and enhancing consumer trust; centralized and physical dual management, data is stored in the central database and RFID tags at the same time, which is convenient for inspection and supply chain tracking. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is a structural diagram of a green vegetable quality safety traceability management system of the present invention.

[0029] Figure 2 This is a structural diagram of a fresh vegetable product quality and safety traceability system in the prior art.

[0030] Figure 3 This is a structural diagram of a vegetable base management system based on traceability technology in the prior art.

[0031] Figure 4 This is a schematic diagram of the login and main interface of the green vegetable quality and safety traceability management system of the present invention.

[0032] Figure 5 This is a schematic diagram of the TPF-CNN model training of the present invention.

[0033] In the figure: Aphids represents aphids; Bollworm represents cotton bollworm; Leaf Folder represents green stink bug; Epochs represents the number of training rounds; Batch Size represents the batch size; Leaning Rate represents the learning rate. DETAILED DESCRIPTION

[0034] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0035] Example 1

[0036] See Figure 1 As shown in the structural diagram, the present invention provides a green vegetable quality and safety traceability management system, which includes a production information collection module, a harvest information collection module, a detection and packaging management module and a quality and safety management module; the production information collection module is used to record and manage key data in the production process starting from the source of green vegetable production; the harvest information collection module uses RFID technology to record and manage data on green vegetables during the harvest process; the detection and packaging management module detects the quality of green vegetables and packages qualified products; the quality and safety management module formulates, implements and monitors production standards in real time. Figure 4 Shown is the login and main interface of the green vegetable quality and safety traceability management system.

[0037] Furthermore, the production information collection module is used to record and manage key data in the production process starting from the source of green vegetable production, including: the production information collection module realizes the automatic collection and management of production data through the application of radio frequency identification (RFID) technology, thereby improving the transparency and traceability of the production process. Farmers first use an RFID reader to scan the greenhouse RFID tag to obtain a unique greenhouse number. The unique greenhouse number serves as a key link between the greenhouse and production data, ensuring the accurate attribution of the data. For example, when a farmer holds the RFID reader close to the tag at the entrance of the greenhouse, the system will automatically identify and display the greenhouse number (such as "GH001"). Next, farmers need to record various data in the production process, including environmental data and production operation record data. The environmental data covers soil fertility (such as nitrogen, phosphorus, and potassium content), humidity (soil moisture percentage), temperature (real-time temperature in the greenhouse), etc. The environmental data directly affects the growth quality of vegetables and is obtained through sensors or manual measurements; the production operation records include fertilization records, pest and disease control measures, and irrigation records. Farmers need to enter detailed information such as the type of fertilizer, dosage, and fertilization time, and enter the specific dosage and date; farmers also need to register the name, use time, and dosage of pesticides to monitor whether the use of chemical substances meets safety standards; finally, farmers record irrigation information, including irrigation time and water consumption.

[0038] After data entry is complete, the environmental data and production operation records are not only stored in the system but also written to the greenhouse's RFID tag. These RFID tags, which use a high-frequency 13.56MHz passive tag, record the greenhouse number and key production data. Simultaneously, all data is uploaded to a MySQL database for centralized management and real-time querying. For example, when a farmer completes a fertilization record, the system automatically writes the data to the tag and transmits the information via the network to the MySQL database on the server. Managers can query the fertilization history of a particular greenhouse at any time through the system interface.

[0039] Furthermore, the harvest information collection module uses RFID technology to record and manage data on the vegetables during the harvest process, including: the harvest information collection module includes three steps for harvesting vegetables, as follows:

[0040] In step Y1, the farmer uses an RFID reader to scan the RFID tag attached to the greenhouse. This RFID tag can be read without direct line of sight, speeding up the scanning process. The RFID tag contains a detailed production history of the vegetables during the growing phase. This history includes the environmental conditions (temperature, humidity, and soil quality) used to grow the vegetables, as well as the fertilizers, pesticides, and irrigation used during the growing period.

[0041] In step Y2, farmers scan the RFID tags on the harvest trays used to collect the vegetables during harvest. Each harvest tray is equipped with a unique RFID tag that serves as an identifier for that particular batch of vegetables. After scanning the tray tag, farmers enter additional harvest details directly into the system. These details include the exact time and date of harvest, the specific picking location within the greenhouse or on the farm, and the quantity of vegetables harvested.

[0042] In step S3, production history is linked to harvest information, creating a record that links the entire lifecycle of the greens. This combined data is then stored in two locations: the harvest pallet's RFID tag and a central MySQL database. Storing information directly on the pallet's RFID tag ensures that the data physically moves with the product through the supply chain. This allows for rapid on-site verification without accessing a central system; for example, a quality inspector can scan a pallet tag and immediately retrieve production and harvest details. Simultaneously, synchronizing data to a central database provides a centralized repository for all harvest records, enabling broader analysis, reporting, and oversight.

[0043] Linking production history with harvest information allows any issues detected later to be traced not only to harvest conditions but also to specific production practices. For example, if a post-harvest inspection reveals excessive pesticide residues in a batch of green vegetables, the system can quickly determine whether the problem stems from improper pesticide application during production or from harvesting too early after application. By automatically collecting data through RFID scanning and direct input, the harvest information collection module reduces the need for manual recordkeeping, which is time-consuming and prone to errors.

[0044] Furthermore, the inspection and packaging management module inspects the quality of green vegetables and packages qualified products, including: first, the inspector uses the RFID reader to scan the RFID tag on the harvest tray, which can quickly obtain the production and harvest history information of the green vegetables, such as the greenhouse number, planting date, fertilization record, pest and disease control measures, and harvest time and location. Then, the inspection and packaging management module inspects the quality of the green vegetables and checks for pests and pesticide residues: the inspection and packaging management module detects pests through the TPF-CNN model; the pesticide content is detected through a fixed-position pesticide residue detector; finally, the inspector packages the green vegetables with pesticide residues below 0.05 mg / kg and without pests. Each batch of green vegetables will be attached with an RFID tag containing traceability information, and the RFID tag stores the entire process data from production, harvesting to inspection. The TPF-CNN model for detecting pests includes the following specific contents:

[0045] In step S1, a camera is placed on the water flow for detecting pests, and the camera is aimed at the vegetables to shoot the captured images, and the captured images are used as input data for subsequent identification.

[0046] Step S2: The captured original image is preprocessed to improve image quality and reduce interference. The preprocessing includes cropping, scaling, and contrast adjustment. The cropping is to remove irrelevant background parts in the image; the scaling is to adjust the image size to meet the input requirements of the TPF-CNN model; and the contrast adjustment is to enhance the visual characteristics of the pests.

[0047] In step S3, the TPF-CNN model performs feature extraction and pest detection on the preprocessed image.

[0048] The TPF-CNN model is a 28-layer convolutional neural network that incorporates a transition probability function (TPF) for pest detection. The TPF first converts continuous features of the input image (such as color gradients and texture changes) into discrete features. Discretization reduces feature complexity, allowing the convolutional neural network to more efficiently detect and classify pests. The specific steps are:

[0049] Step W1: extract the continuous features to be processed from the image.

[0050] In step W2, the range of continuous features is divided into several discrete states. For color values, the grayscale range of 0-255 is divided into 10 intervals, each representing a state. For texture features, different states are defined based on the range of gradient intensity, such as "smooth," "moderate change," and "dramatic change."

[0051] Step W3: For adjacent pixels or regions in the image, calculate the transition probability between their characteristic states. If the grayscale value of a pixel belongs to state A (0-25) and the grayscale value of its adjacent pixel to the right belongs to state B (26-50), then record a transition from A to B. Count the state transition frequencies of all adjacent pixel pairs in the entire image and normalize them into probability values. The calculation formula for the transition probability is: P ij (t)=P{X(t+s)=j|X(s)=i}, where P ij (t) represents the probability of being in state j after time t when being in state i at time s; X(t+s)=j represents the state j at time t+s; X(s)=i represents the state i at time s; P represents the probability symbol.

[0052] Step W4: Use the calculated transition probability as a new feature representation.

[0053] The first 20 layers of the TPF-CNN model consist of 10 groups of "convolutional layers and pooling layers", each group containing one convolutional layer and one pooling layer. The convolution kernel size of each convolutional layer is 3x3, with a stride of 1, and the size of each pooling layer is 2x2, with a stride of 2. The middle 5 layers of the TPF-CNN model are all convolutional layers with a convolution kernel size of 3x3 and a stride of 1. The last 3 layers of the TPF-CNN model consist of 2 fully connected layers and 1 output layer. The input of the first fully connected layer is 512 one-dimensional vectors, and the output size is 4096 one-dimensional vectors. The input of the second fully connected layer is 4096 one-dimensional vectors, and the output is 4096 one-dimensional vectors. The input of the output layer is 4096 one-dimensional vectors, and the output is 6 one-dimensional vectors, of which 5 one-dimensional vectors correspond to 5 pest categories. The five pests include "aphids", "cotton bollworms", "green stink bugs", "leaf rollers" and "leaf miners", among which one one-dimensional vector corresponds to "no pests".

[0054] The training process of the TPF-CNN model is based on a dataset containing 600 images, including five types of pests: "aphids", "cotton bollworms", "green stink bugs", "leaf rollers" and "leaf miners", with 100 images of each pest and 100 images of no pests. Figure 5 As shown in the figure, the training platform used was Google Teachable Machine, the number of training epochs was set to 100, and the batch size was set to 32, meaning that each training session processed the features and labels of 32 images. The learning rate was set to 0.001.

[0055] Furthermore, the quality and safety management module establishes, implements, and monitors production standards in real time. This module includes the following specific components: The quality and safety management module consists of three steps, ensuring the high quality and safety of green vegetable products through standardized management combined with real-time monitoring. The first step involves technical personnel developing detailed production, harvesting, and testing standards based on the latest national and industry standards. These standards include specific greenhouse environmental requirements, such as a temperature of 20-25 degrees Celsius, humidity of 60-70%, and light conditions that meet crop growth requirements, ensuring optimal growth. They also define pesticide and fertilizer usage limits, such as the dosage and safe intervals for each pesticide, to prevent excessive chemical residues. Furthermore, specific procedures for harvest timing and packaging are included. The second step involves production managers rigorously reviewing the standards developed by technical personnel, focusing on their scientific validity, practicality, and operability. During the review, managers will consider production costs, available equipment, and staff skill levels to ensure the standards are practical and achievable. Once the standard is approved, administrators will disseminate it to all stages of the production, harvesting, and testing process. Training will be organized to ensure farmers and testers fully understand and master the standard, and the system's operational guidelines and checklists will be updated. The third step involves the system monitoring the entire production process in real time. Using sensors and RFID technology, the system automatically collects greenhouse environmental data, pesticide usage records, and harvest information, comparing it to pre-set standards in real time. If an anomaly is detected, such as a greenhouse temperature exceeding 25°C or excessive pesticide usage, the system will immediately issue an alarm, notifying management and operators to take timely corrective measures, such as turning on ventilation or ceasing spraying.

[0056] The present invention uses RFID tags to record greenhouse environmental data (such as temperature, humidity, soil quality) and production operation information (such as fertilizer and pesticide use), and links production history with harvest details to ensure data transparency and traceability, overcoming the limitations of existing systems that cannot accurately locate production units and are inefficient. In addition, the system integrates a TPF-CNN model for pest detection to further ensure the health of vegetables and reduce errors and time-consuming manual records. Its technical effects and advantages include: automation and efficiency, using RFID to achieve automatic data collection, suitable for large-scale production; accurate traceability, which can quickly trace problems to specific production links; quality control, the CNN model improves the accuracy of pest and disease detection; real-time monitoring and transparency, supporting real-time management of the production process and enhancing consumer trust; centralized and physical dual management, data is stored in the central database and RFID tags at the same time, facilitating inspection and supply chain tracking.

[0057] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0058] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A green vegetable quality safety traceability management system, characterized in that: It includes a production information collection module, a harvest information collection module, a testing and packaging management module, and a quality and safety management module; the production information collection module is used to record and manage key data in the production process starting from the source of vegetable production; The harvest information collection module uses RFID technology to record and manage data on the harvest process of green vegetables; the inspection and packaging management module inspects the quality of green vegetables and packages qualified products; the quality and safety management module formulates, implements and monitors production standards in real time; In the inspection and packaging management module, inspectors use RFID readers to scan RFID tags on harvest trays to obtain production and harvest history information for the vegetables. The module then performs quality inspections on the vegetables, checking for pests and pesticide residues. Pests are detected using the TPF-CNN model, and pesticide residue detectors are used to measure pesticide levels. Finally, inspectors package vegetables with pesticide residues below 0.05 mg / kg and those free of pests. Each batch of vegetables is affixed with an RFID tag containing traceability information, which stores data from production, harvesting, and inspection. The TPF-CNN model detects pests using the following specific procedures: Step S1, photographing the vegetables with a camera to obtain vegetable image data; Step S2, preprocessing the captured original image to improve image quality and reduce interference, wherein the preprocessing includes cropping, scaling, and contrast adjustment; In step S3, the TPF-CNN model performs feature extraction and pest detection on the preprocessed image.

2. A vegetable quality safety traceability management system according to claim 1, characterized in that The TPF-CNN model is a 28-layer convolutional neural network combined with a transition probability function for detecting pests. The transition probability function first converts the continuous features of the input image into discrete features. Discretization reduces the complexity of the features so that the convolutional neural network can detect and classify pests more efficiently. The specific steps are as follows: Step W1, extracting the continuous features to be processed from the image; Step W2: Divide the value range of the continuous feature into several discrete states. For color values, the grayscale range of 0-255 is divided into 10 intervals, each representing a state. For texture features, different states are defined based on the range of gradient intensity. Step W3: For adjacent pixels or regions in the image, calculate the transition probability between their characteristic states. If the grayscale value of a pixel belongs to state A and the grayscale value of its adjacent pixel to the right belongs to state B, then record a transition from A to B. Count the state transition frequencies of all adjacent pixel pairs in the entire image and normalize them to probability values. Step W4: Use the calculated transition probability as a new feature representation.

3. A green vegetable quality safety traceability management system according to claim 1, characterized in that The first 20 layers of the TPF-CNN model are composed of 10 groups of "convolutional layers and pooling layers", each group contains a convolutional layer and a pooling layer; the convolution kernel size of each convolutional layer is 3x3, the stride is 1, and the size of each pooling layer is 2x2, the stride is 2; the middle 5 layers of the TPF-CNN model are all convolutional layers with a convolution kernel size of 3x3 and a stride of 1; the last 3 layers of the TPF-CNN model are composed of 2 fully connected layers and 1 output layer, the input of the first fully connected layer is 512 one-dimensional vectors, and the output size is 4096 one-dimensional vectors; the input of the second fully connected layer is 4096 one-dimensional vectors, and the output is 4096 one-dimensional vectors; the input of the output layer is 4096 one-dimensional vectors, and the output is 6 one-dimensional vectors, of which 5 one-dimensional vectors correspond to 5 pest category labels, and the last one-dimensional vector corresponds to the "no pest" label.

4. A green vegetable quality safety traceability management system according to claim 1, characterized in that: The cropping is to remove irrelevant background parts in the image; the scaling is to adjust the image size to meet the input requirements of the TPF-CNN model; and the contrast adjustment is to enhance the visual features of pests.

5. A vegetable quality and safety traceability management system according to claim 1, wherein the production information collection module collects data using RFID technology. Farmers use RFID readers to scan greenhouse RFID tags to obtain unique numbers, record environmental data and production operation records, and write the data into high-frequency 13.56MHz passive tags and upload them synchronously to a MySQL database.

6. The vegetable quality and safety traceability management system according to claim 1, wherein the harvest information collection module includes three steps: scanning greenhouse RFID tags to obtain production history, scanning harvest tray RFID tags and entering harvest details, and associating production history with harvest information and storing them on tray tags and in a central database.

7. A green vegetable quality and safety traceability management system according to claim 1, wherein the production history includes planting environment conditions and fertilizer, pesticide, and irrigation records, and the harvest pallet label serves as a batch identifier.

8. A green vegetable quality and safety traceability management system according to claim 1, wherein the quality and safety management module includes three steps, which ensure the high quality and safety of green vegetable products by combining standardized management with real-time monitoring. The first step of the quality and safety management module is for technical personnel to formulate production, harvesting, and testing standards, including greenhouse temperature of 20-25°C, humidity of 60-70%, pesticide dosage and safety interval, harvesting time and packaging procedures; the second step of the quality and safety management module is for production managers to review the scientificity and practicality of the standards, distribute them to each link and organize training, and update the system operation guide and checklist; the third step of the quality and safety management module is for the system to automatically collect data through sensors and RFID, compare it with preset standards, and immediately issue an alarm when an abnormality occurs.

9. A vegetable quality and safety traceability management system according to claim 1, wherein the greenhouse RFID tag records the greenhouse number and key production data, and the MySQL database realizes centralized data management and real-time query. After the farmer enters the data, the system automatically writes the tag and uploads it to the database.

Images