Meat product safety tracing and whole-process monitoring system and method

Abstract

The invention provides a meat product safety tracing and whole-process monitoring system and method, belongs to the field of food safety, and can comprehensively monitor the whole process of meat products from raw materials to terminal retail in real time. Meanwhile, a quality and risk prediction model is also constructed, the model is trained, and the purposes of timely early warning, prevention and control are achieved by detecting parameters such as temperature, humidity and the like of a field link, so that the food safety is ensured. According to the method, the reliability and transparency of data are ensured by adopting a block chain technology, and automatic management is realized by utilizing an intelligent contract. A consumer can check the whole-course information of the meat product through the AR / VR query system. The whole-course traceability of the meat products from the farm to the dining table is realized, so that consumers can eat the meat products at ease, and the trust of the consumers is won for enterprises. The popularization and application of the method can play a positive role in promoting the healthy development of the food industry.

Description

Technical Field

[0001] The present invention belongs to the field of food safety, and more specifically relates to a meat product safety tracing and full-process monitoring system and method. Background Art

[0002] As an important part of the global food industry, the meat products industry faces severe challenges in product safety supervision. Previously, safety issues of meat products frequently occurred, which seriously damaged the trust of consumers and the image of the industry. The causes of these problems often come from problems at the source of the product, lack of supervision in the production process, and potential safety hazards in the sales process. Therefore, it is particularly important to effectively monitor the entire process of meat products.

[0003] Among them, there are two main problems: one is the traceability problem, and the other is the safety monitoring problem.

[0004] Regarding the first problem, traditional traceability methods rely on manual registration and review, which is time-consuming, labor-intensive, and prone to errors, and cannot meet the needs of fast and accurate traceability. Although the introduction of automation technology has improved this problem, there are still difficulties in traceability due to configuration and algorithm issues in a single system. The introduction of blockchain technology, with its tamper-proof characteristics, can effectively achieve traceability, but the common problem is that this method cannot correspond to physical objects, and abuse is also common.

[0005] Regarding the second problem, traditional safety monitoring relies on manual inspection and visual identification, and is powerless against microorganisms, chemical substances, and other problems that cannot be identified by the naked eye. Although some companies have introduced automated equipment for testing, they still cannot monitor the entire process. The introduction of IoT technology has made this problem possible, but the common problem is that the collected data is scattered and difficult to integrate, and effective mining and utilization of this data is very limited. Summary of the invention

[0006] In view of the above problems, the present invention proposes a method for safety traceability and full-process monitoring of meat products. By combining blockchain technology and Internet of Things technology, corresponding data collection parameters and quality as well as risk prediction models are set to achieve safety traceability and full-process monitoring of meat products.

[0007] In order to achieve the above object, the present invention adopts the following technical solution: the monitoring method comprises:

[0008] Choose blockchain for deployment and issue digital tokens to mark each meat product;

[0009] Configure IoT devices: Deploy appropriate IoT devices such as RFID and sensors at all stages of meat production, and set corresponding data collection parameters.

[0010] Establish quality and risk prediction models: Use historical and existing data to establish quality and risk prediction models;

[0011] Training quality and risk prediction models;

[0012] In one embodiment, the blockchain is deployed including:

[0013] Choose Ethereum as the blockchain platform, configure the network structure of each node, and the permission mechanism of the blockchain network;

[0014] Create digital tokens through smart contracts and use the ERC-20 standard to issue digital tokens;

[0015] Encrypt digital tokens.

[0016] In one embodiment, the configured IoT device includes an RFID tag, a reader, and a sensor;

[0017] RFID reading devices are installed on production lines, warehouse entrances and exits, and retail outlets; temperature and humidity sensors are installed in the production, storage, and transportation environments of meat products;

[0018] The RFID reader sets the reading frequency; the temperature and humidity sensors set the data reporting cycle and threshold;

[0019] Connect all IoT devices to a central system that is responsible for collecting, storing, and processing the data collected by the devices.

[0020] In one embodiment, the quality and risk prediction model is established by using a graph-based semi-supervised learning method, including:

[0021] (1) Constructing a graph model:

[0022] Map the meat product data into a graph where each data point is a node;

[0023] There are n samples in the data set D, of which l samples are labeled and u samples are unlabeled, so n = l + u;

[0024] Each node v i (1≤i≤n) represents a sample x i ; If the sample x i and sample x j is similar, there is a line from node v in the figure i To node v j The edge ij ;

[0025] This graph is represented by an n×n matrix W, where W ij Represents node v i and node v j The similarity is calculated as follows:

[0026]

[0027] (2) Set the label:

[0028] For labeled data points, set it to l, and for unlabeled data points, set it to u; in the meat product traceability problem, these labels are qualified or unqualified;

[0029] (3) Use the first-order approximation algorithm to generate new labels:

[0030] Define f as a function on the graph G, for a given vertex label f l , predict the label f ​​of the unlabeled vertex u ; It is expressed as the problem of minimizing the loss function, and the approximation method is defined as H = (I-αW) -1 Y, where α is the empirically selected damping coefficient and Y is the label matrix;

[0031] (4) Forecasting principles:

[0032] The principle of label prediction for unlabeled data points is: if most of the "neighbors" of a data point, the other points connected to it in the graph, are labeled as a certain category, then this data point will also be labeled as this category.

[0033] In one embodiment, the training of the quality and risk prediction model includes:

[0034] Initialize labels: For labeled data points, initialize their labels y first; use matrix F to represent the labels of all samples, which is an n×c matrix, where c is the number of categories; for labeled samples x i , and set the corresponding position in F to the label value: f i =y i ; For unlabeled samples, first set their labels in F to the mean of all possible values, that is,

[0035] Iterative optimization of labels: After obtaining the initial labels, the iterative optimization process begins, using label propagation, where the label propagation formula is as follows:

[0036] F (t+1) =αWF (t) +(1-α)Y

[0037] In the above formula, F (t)represents the label matrix after the tth iteration, Y is the initial label matrix, and α is an adjustable parameter used to determine the weight of the two;

[0038] Voting to select labels: After the iteration, the predicted labels of all samples will be obtained, and the final label will be determined by voting. That is, the label corresponding to the column with the largest value will be considered as the final label of the sample.

[0039] In another aspect, a meat product safety traceability and full-process monitoring system is provided, wherein the monitoring system is applicable to the method described above, and the monitoring system comprises:

[0040] The database is responsible for storing and managing all generated data, including biological tags of meat products, real-time data of parameters at each stage, and quality and risk prediction models;

[0041] The IoT device module includes various IoT device RFID tags, reading devices, and sensors, which are responsible for real-time monitoring of production lines, warehouses, and retail outlets, and transmitting data to the database;

[0042] The blockchain module is responsible for deploying and managing the blockchain platform, issuing digital tokens to mark each meat product, and storing data on the chain;

[0043] The bio-label production module is responsible for generating the bio-label of meat products and storing the bio-label information in the blockchain and database;

[0044] AR / VR query system module, providing AR / VR interface, scanning QR code on meat products, and viewing specific information of each link of meat products through AR / VR application;

[0045] The quality and risk prediction model module predicts and detects existing risks in real time based on the collected data and preset quality parameter thresholds;

[0046] Alert system module When the quality and risk prediction model finds problems with meat products, an alarm is triggered and marked on the blockchain.

[0047] Beneficial effects of the present invention:

[0048] The present invention provides a method and system for meat product safety traceability and full-process monitoring, which effectively combines blockchain technology and Internet of Things devices to achieve efficient traceability and monitoring of the entire process of meat products. The authenticity and accuracy of the data are ensured, and the various safety parameters and link status of meat products can be monitored in real time. By constructing and training quality and risk prediction models, timely and accurate risk warnings can be obtained, so that safety issues can be discovered and dealt with as early as possible to ensure food safety. At the same time, through the application of blockchain and AR / VR technology, the transparency of meat products is improved, and consumers' trust in the safety of meat products is enhanced. Finally, the method and system improve the efficiency of meat product safety traceability and full-process monitoring, and have a positive role in promoting the development of the entire meat product industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 is a flow chart of the method of the present invention;

[0050] Figure 2 Establish a quality and risk prediction model flow chart for the present invention;

[0051] Figure 3 Establishing a quality and risk prediction model training flow chart for the present invention;

[0052] Figure 4 This is a diagram of the meat product safety tracing and full-process monitoring system of the present invention. DETAILED DESCRIPTION

[0053] In order to facilitate the understanding of the present invention, the present invention will be described more fully below with reference to the relevant drawings. Typical embodiments of the present invention are given in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described in the present invention. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive.

[0054] Unless otherwise defined, all technical and scientific terms used in the present invention have the same meaning as those understood by those skilled in the art of the present invention. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0055] Embodiment 1:

[0056] like Figure 1 As shown, the meat product safety traceability and full-process monitoring method comprises:

[0057] S1. Select blockchain for deployment and issue digital tokens to mark each meat product.

[0058] The blockchain deployment includes:

[0059] Choose the blockchain platform Ethereum, which supports smart contracts and is suitable for creating complex applications.

[0060] Configure the blockchain network: You need to configure the network structure of each node, as well as the permission mechanism of the blockchain network, whether anyone is allowed to join the network, or only authorized persons can join.

[0061] Creating digital tokens: This is done through smart contracts. On Ethereum, you can use the ERC-20 standard it provides to issue your own tokens.

[0062] This process involves encryption and decryption. Public key cryptography is the core of blockchain technology, which includes the following two formulas:

[0063] Signature: Sig = sign (SK, message)

[0064] Verify: verify(PK, Sig, message) == True

[0065] Among them, SK is the private key, PK is the public key, message is the message, sign is the signature function, and verify is the verification function. The signing process generates a signature through the private key and the message, and the verification process verifies whether the signature is valid through the public key, signature, and message.

[0066] S2. Configure IoT devices: Deploy appropriate IoT devices such as RFID and sensors at all stages of meat production, and set corresponding data collection parameters.

[0067] First, it is necessary to select IoT devices suitable for deployment in various links of meat production, processing, transportation and sales. These devices include RFID tags, reading devices, sensors, etc. RFID tags serve to track the movement of meat products in the supply chain, and temperature sensors are used to monitor the temperature in the storage and transportation links.

[0068] Equipment needs to be deployed in the right locations to collect key data of meat products marked with digital tokens. RFID reading equipment needs to be installed in production lines, warehouse entrances and exits, retail outlets, etc. Temperature and humidity sensors need to be installed in the production, storage and transportation environments of meat products.

[0069] After the equipment is configured, the relevant data collection parameters need to be set. The RFID reader needs to set the reading frequency; the temperature and humidity sensors need to set the data reporting cycle and threshold, etc. These parameters should be set based on the safety requirements of meat products and the characteristics of the supply chain.

[0070] Connect all IoT devices to a central system that collects, stores and processes the data collected by the devices, generates real-time reports and issues warnings when quality issues arise.

[0071] S3. Establish quality and risk prediction models: Use historical and existing data to establish quality and risk prediction models.

[0072] First, we need to use big data technology to collect, clean, and integrate various historical and existing meat product data, which can include multi-dimensional data such as production date, origin, storage environment, transportation environment, and sales information. Then, based on this data, combined with the semi-supervised learning method of the graph, a prediction model is established to predict the quality of meat products and possible risks.

[0073] like Figure 2 As shown, the quality and risk prediction model is established by using a graph-based semi-supervised learning method, including:

[0074] S301. Build a graph model:

[0075] Map the meat product data into a graph where each data point is a node;

[0076] There are n samples in the data set D, of which l samples are labeled and u samples are unlabeled, so n = l + u;

[0077] Each node v i (1≤i≤n) represents a sample x i ; If the sample x i and sample x j is similar, there is a line from node v in the figure i To node v j The edge ij ;

[0078] This graph is represented by an n×n matrix W, where W ij Represents node v i and node v j The similarity is calculated as follows:

[0079]

[0080] S302, set label:

[0081] For labeled data points, set it to l, and for unlabeled data points, set it to u; in the meat product traceability problem, these labels are qualified or unqualified;

[0082] S303, using the first-order approximation algorithm to generate a new label:

[0083] Define f as a function on the graph G, for a given vertex label f l , predict the label f ​​of the unlabeled vertex u ; It is expressed as the problem of minimizing the loss function, and the approximation method is defined as H = (I-αW) -1 Y, where α is the empirically selected damping coefficient and Y is the label matrix;

[0084] S304. Prediction principles:

[0085] The principle of label prediction for unlabeled data points is: if most of the "neighbors" of a data point, the other points connected to it in the graph, are labeled as a certain category, then this data point will also be labeled as this category.

[0086] Using unlabeled data for model training: In the whole process monitoring of meat product safety, a large amount of data with unpredictable labels will be generated, such as various environmental parameters. Through semi-supervised learning, these unlabeled data can be used for model training to improve the accuracy of prediction.

[0087] Reduce training costs: Since semi-supervised learning can use unlabeled data for training, it can greatly reduce the cost and time of manually annotating data.

[0088] Improve the generalization ability of the model: Graph-based semi-supervised learning takes into account the distribution and structure information of the data during the prediction process, so that the learned model has better generalization ability.

[0089] Processing high-dimensional data: In the process of meat product traceability, it is necessary to monitor multiple links and collect multiple parameters, resulting in high data dimensions. Graph-based semi-supervised learning can handle this kind of high-dimensional data very well. Since each node of the graph represents a sample and the edges between nodes represent the similarities between samples, the data structure can be concisely and clearly represented by the graph model regardless of the data dimension.

[0090] The judgment principle is simple and clear: if most of the "neighbors" of a data point, the other points connected to it in the graph, are marked as a certain category, then this data point will also be marked as this category. This judgment principle is concise and intuitive and easy to operate.

[0091] In addition, the data generated during the whole process of meat product safety monitoring is extremely rich in form and has significant structure. These data include both scalar data such as temperature and humidity, and high-dimensional spatial data such as images and video streams. The graph-based semi-supervised learning method can effectively process these data, mine the intrinsic correlation of the data, and perform model prediction, greatly improving the accuracy and efficiency of meat product traceability and whole-process monitoring.

[0092] S4. Train the quality and risk prediction model. After the model is established, it needs to be trained. This is the process of training the quality and risk prediction model. Through continuous iterative training, a model that can accurately predict the quality of future meat products and possible risks is finally generated.

[0093] like Figure 3 As shown, the training of the quality and risk prediction model includes:

[0094] S401, initialize labels: For the labeled data points, initialize their labels y first; use matrix F to represent the labels of all samples, which is an n×c matrix, c is the number of categories; for the labeled samples x i , and set the corresponding position in F to the label value: f i =y i ; For unlabeled samples, first set their labels in F to the mean of all possible values, that is,

[0095] S402, iterative optimization of labels: After obtaining the initial labels, the iterative optimization process begins, using label propagation, where the label propagation formula is as follows:

[0096] F (t+1) =αWF (t) +(1-α)Y

[0097] In the above formula, F (t) represents the label matrix after the tth iteration, Y is the initial label matrix, and α is an adjustable parameter used to determine the weight of the two;

[0098] S403, voting to select labels: After the iteration is completed, the predicted labels of all samples are obtained, and the final labels are determined by voting, that is, the label corresponding to the column with the largest value is considered to be the final label of the sample.

[0099] S5. Conduct actual meat product safety prediction. After the model is established and trained, meat product safety prediction can be conducted. This prediction process mainly includes inputting various newly collected data into the trained model. The model calculates and judges based on these data and outputs the prediction results of meat product quality and risk.

[0100] This prediction result can be used as reference information to guide decision-making in various links such as production, transportation, storage and sales of meat products, so as to avoid potential quality problems and risks, ensure the overall quality and safety of meat products, and provide a basis for regulatory authorities.

[0101] If the model predicts that there are quality problems or potential risks, an alarm will be triggered and marked on the blockchain, so that any stakeholders, including consumers, can get this information from the blockchain and understand the true situation of the meat products. Therefore, this model is crucial for the entire meat product safety tracking and full-process monitoring system.

[0102] Embodiment 2:

[0103] like Figure 4 As shown, a meat product safety traceability and full-process monitoring system, the monitoring system is applicable to the method described, and the monitoring system includes:

[0104] The database is responsible for storing and managing all generated data, including biological tags of meat products, real-time data of parameters at each stage, and quality and risk prediction models; the main function of this module is to store and maintain data. In the specific implementation process, a database system is selected, such as MySQL. Then, a table structure or document structure is defined to store different types of data (such as biological tags of meat products, real-time data of various production links, quality and risk prediction models, etc.).

[0105] The IoT device module includes various IoT devices such as RFID tags, readers, and sensors, which are responsible for real-time monitoring of production lines, warehouses, and retail outlets, and transmitting data to the database. This module contains many IoT devices such as RFID tags, readers, sensors, etc. In the implementation process, each device is deployed in different links of meat products, such as production lines, warehouses, and retail outlets, according to demand. These devices continuously collect data from the environment and then send the data to the database through a (wireless) network protocol (Wi-Fi, cellular network, Bluetooth, etc.).

[0106] The blockchain module is responsible for deploying and managing the blockchain platform, issuing digital tokens to mark each meat product, and storing the data on the chain. Smart contracts need to be deployed on the selected blockchain platform, such as Ethereum. Then, by calling the smart contract, digital tokens are created and assigned to each meat product. The blockchain module is also responsible for putting the hash value of the data on the chain to ensure that the data cannot be tampered with.

[0107] The bio-label production module is responsible for generating the bio-label of meat products and storing the bio-label information in the blockchain and database. The bio-label production process is as follows:

[0108] Conduct nutritional analysis on meat products. This process may include determining the content of nutrients such as protein, fat, carbohydrates, minerals, vitamins, etc. Specific nutritional analysis methods can refer to standard methods related to food analysis.

[0109] Generate biological labels: Generate biological labels for meat products using the above extracted nutritional analysis results. This biological label is a hash value that can be represented by the following hash function:

[0110] H(dna,nutrition)=hash

[0111] Among them, H represents the hash function, DNA and nutrition represent DNA information and nutritional analysis results respectively, and hash represents the generated biological label.

[0112] Attach biological labels to meat products: Attach biological labels to the packaging of meat products in the form of QR codes or other forms. In the subsequent traceability process, the biological labels of meat products can be obtained by scanning the QR codes, thereby obtaining the DNA information and nutritional analysis results of the meat products.

[0113] The AR / VR query system module provides an AR / VR interface, scans the QR code on the meat products, and views the specific information of each link of the meat products through the AR / VR application.

[0114] First, develop an AR / VR application. This is done by using AR / VR development tools such as Unity or UnrealEngine. Important functions include, but are not limited to, scanning QR codes and presenting data.

[0115] The application needs to be able to connect to the database to access the stored meat product information. This usually involves writing a data interface such as an API.

[0116] Visualize the acquired meat product data in an AR / VR environment. This may involve 3D modeling, animation design, etc. The goal of data visualization is to simplify and intuitively display complex data information as much as possible.

[0117] The quality and risk prediction model module predicts and detects existing risks in real time based on the collected data and preset quality parameter thresholds.

[0118] Alert system module When the quality and risk prediction model finds problems with meat products, an alarm is triggered and marked on the blockchain.

[0119] Define the events and parameters that need to trigger alarms in the system.

[0120] When an alarm event is triggered, set how the system should notify relevant personnel. Possible ways include: sending emails, text messages, push notifications, etc. When the predictive model detects a possible risk, the meat product should be marked on the blockchain at the same time to ensure the transparency and immutability of the alarm information.

[0121] Embodiment three:

[0122] Assume that a batch of pork products is produced in a farm in Shandong, and the farm uses the monitoring system proposed by the present invention to monitor the whole process of safety. Pork products include slaughtering, cutting, packaging and transportation of live pigs.

[0123] When the pigs enter the production line through the RFID reading device, the RFID reading device will read the RFID tag marked with the digital token, record the data and upload it to the blockchain and database. At this time, the temperature and humidity sensors of the production line begin to monitor the environmental parameters. The thresholds of temperature and humidity are set at 18-24℃ and 50-60% humidity.

[0124] When the slaughter is completed, the pork is divided into 100 portions, each of which is represented by an independent digital token. These digital tokens will be issued on the Ethereum blockchain using smart contracts by the blockchain module. The encryption of digital tokens ensures its security.

[0125] Next, 100 portions of pork are packed and stored in a warehouse. The warehouse is also equipped with temperature and humidity sensors, with parameter thresholds set at 2-7 degrees and 50-55% humidity. If the data at any moment exceeds this range, the system will automatically trigger an alarm.

[0126] Ideally, after a period of time, these 100 portions of pork are shipped to retail stores. The retail stores are also equipped with RFID reading devices. When the pork arrives, the RFID reading device automatically reads the digital token information in the tag and uploads this data to the database and blockchain.

[0127] After consumers purchase one of the portions of pork, they can use the AR / VR application on their mobile phone to scan the QR code on the pork packaging to view specific information about the production, storage, and transportation of the pork.

[0128] In addition, the predictive model will continue to maintain and learn new data to improve the accuracy of future predictions. The alert system can help prevent problematic meat products from entering the market and protect consumers' food safety.

[0129] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM).

[0130] It should be understood that the detailed description of the technical solutions of the present invention by means of the preferred embodiments is illustrative rather than restrictive. A person skilled in the art may modify the technical solutions described in the embodiments, or replace some of the technical features by equivalents, based on reading the specification of the present invention; and these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A meat product safety traceability and full-process monitoring method, characterized in that: The monitoring method comprises: Choose blockchain for deployment and issue digital tokens to mark each meat product; Configure IoT devices: deploy RFID and sensor-suitable IoT devices at all stages of meat products, and set corresponding data collection parameters; Establish quality and risk prediction models: Use historical and existing data to establish quality and risk prediction models; Train quality and risk prediction models.

2. A meat product safety traceability and full-process monitoring method according to claim 1, characterized in that: The blockchain deployment includes: Choose Ethereum as the blockchain platform, configure the network structure of each node, and the permission mechanism of the blockchain network; Create digital tokens through smart contracts and use the ERC-20 standard to issue digital tokens; Encrypt digital tokens.

3. The meat product safety traceability and full-process monitoring method according to claim 1, characterized in that: The configured IoT devices include RFID tags, reading devices, and sensors; RFID reading devices are installed on production lines, warehouse entrances and exits, and retail outlets; temperature and humidity sensors are installed in the production, storage, and transportation environments of meat products; The RFID reader sets the reading frequency; the temperature and humidity sensors set the data reporting cycle and threshold; Connect all IoT devices to a central system that is responsible for collecting, storing, and processing the data collected by the devices.

4. The meat product safety traceability and full-process monitoring method according to claim 1, characterized in that: The quality and risk prediction model is established by using a graph-based semi-supervised learning method, including: (1) Constructing a graph model: Map the meat product data into a graph where each data point is a node; There are n samples in the data set D, of which l samples are labeled and u samples are unlabeled, so n = l + u; Each node v i (1≤i≤n) represents a sample x i ; If the sample x i and sample x j is similar, there is a line from node v in the figure i To node v j The edge ij ; This graph is represented by an n×n matrix W, where W ij Represents node v i and node v j The similarity is calculated as follows: (2) Set the label: For labeled data points, set it to l, and for unlabeled data points, set it to u; in the meat product traceability problem, these labels are qualified or unqualified; (3) Use the first-order approximation algorithm to generate new labels: Define f as a function on the graph G, for a given vertex label f l , predict the label f ​​of the unlabeled vertex u ; It is expressed as the problem of minimizing the loss function, and the approximation method is defined as H = (I-αW) -1 Y, where α is the empirically selected damping coefficient and Y is the label matrix; (4) Forecasting principles: The principle of label prediction for unlabeled data points is: if most of the "neighbors" of a data point, the other points connected to it in the graph, are labeled as a certain category, then this data point will also be labeled as this category.

5. The meat product safety traceability and full-process monitoring method according to claim 1, characterized in that: The training of the quality and risk prediction model includes: Initialize labels: For labeled data points, initialize their labels y first; use matrix F to represent the labels of all samples, which is an n×c matrix, where c is the number of categories; for labeled samples x i , and set the corresponding position in F to the label value: f i =y i ; For unlabeled samples, first set their labels in F to the mean of all possible values, that is, Iterative optimization of labels: After obtaining the initial labels, the iterative optimization process begins, using label propagation, where the label propagation formula is as follows: F (t+1) =αWF (t) +(1-α)Y In the above formula, F (t) represents the label matrix after the tth iteration, Y is the initial label matrix, and α is an adjustable parameter used to determine the weight of the two; Voting to select labels: After the iteration, the predicted labels of all samples will be obtained, and the final label will be determined by voting. That is, the label corresponding to the column with the largest value will be considered as the final label of the sample.

6. A meat product safety traceability and full-process monitoring system, the monitoring system being applicable to the method according to any one of claims 1 to 5, characterized in that: The monitoring system comprises: The database is responsible for storing and managing all generated data, including biological tags of meat products, real-time data of parameters at each stage, and quality and risk prediction models; The IoT device module includes various IoT device RFID tags, reading devices, and sensors, which are responsible for real-time monitoring of production lines, warehouses, and retail outlets, and transmitting data to the database; The blockchain module is responsible for deploying and managing the blockchain platform, issuing digital tokens to mark each meat product, and storing data on the chain; The bio-label production module is responsible for generating the bio-label of meat products and storing the bio-label information in the blockchain and database; AR / VR query system module, which provides AR / VR interface, scans QR code on meat products, and uses AR / VR application to view specific information of each link of meat products; The quality and risk prediction model module predicts and detects existing risks in real time based on the collected data and preset quality parameter thresholds; Alert system module When the quality and risk prediction model finds problems with meat products, an alarm is triggered and marked on the blockchain.

Images