Method for detecting and tracing potential hazard factors of cured pork products

Abstract

The invention provides a method for detecting and tracing potential hazard factors of cured pork products, which comprises the following steps of: acquiring multi-dimensional data in production and storage links of the cured pork products, generating corresponding batch information, monitoring real-time states of the cured pork products, and binding the real-time states, the multi-dimensional data and the batch information; the mutually bound batch information, real-time state and multi-dimensional data are uploaded to a block chain, a first smart contract is deployed in the block chain, the multi-dimensional data of each batch of cured pork products are analyzed through the first smart contract, a potential hazard factor detection result is obtained, and the potential hazard factor detection result is sent to the block chain. Writing a potential hazard factor detection result into the block chain and binding the potential hazard factor detection result with batch information; a PUF label is added to the cured pork product, and the PUF label is bound with the batch information; and identifying the PUF label, querying batch information according to an identification result, and further obtaining a corresponding potential hazard factor detection result. The traceability accuracy and the anti-counterfeiting capability can be improved.

Description

Technical Field

[0001] This invention relates to the field of food safety traceability technology, and in particular to a method for detecting and tracing potential hazard factors in cured pork products. Background Technology

[0002] Cured pork products are pork products made from raw meat through pretreatment, curing (sauce-making), and sun-drying (or baking, smoking). They are a popular category of processed meat products with high processing efficiency. However, cured pork product manufacturers are generally small-scale, resulting in low industry concentration, outdated processing techniques, unstable output and quality, a lack of advanced production technology support, and a lack of traceability systems and technologies for hazardous substances. The safety risks associated with cured pork products are complex and numerous. Identification and analysis of potential chemical hazards are crucial for hazard tracing, improving green storage and transportation technologies, and establishing a quality assurance system for cured pork products, thereby promoting the healthy and sustainable development of the pig industry. Therefore, the safety traceability of cured pork products is vital to the survival and development of pork product enterprises.

[0003] "Meat quality traceability" refers to meat products produced from certified farms and identified animals, with documentation detailing their origin and processing. This documentation includes the animal source (batch), the farm that raised the animal, and the slaughterhouse that slaughtered the animal to produce the meat. Current research on meat product traceability technologies focuses on labeling, isotope tracing, organic matter tracing, and DNA tracing. Existing traceability technologies, such as QR codes, often only record basic information like batch number and manufacturer. However, the production process and storage methods of cured pork products can influence the generation of potential chemical hazards, and current traceability technologies cannot detect and trace these potential hazard factors. Summary of the Invention

[0004] The purpose of this invention is to provide a method for detecting and tracing potential hazards in cured pork products. Based on multi-dimensional data from the production and storage of cured pork products, this method combines blockchain technology and PUF tags to achieve the detection and tracing of potential hazards.

[0005] To achieve the above-mentioned objective, this invention provides a method for detecting and tracing potential hazard factors in cured pork products, the method comprising: S101. Collect multi-dimensional data in the production and storage of cured pork products, generate corresponding batch information, monitor the real-time status of cured pork products, and bind the real-time status, multi-dimensional data and batch information. S102. Upload the mutually bound batch information, real-time status, and multi-dimensional data to the blockchain, deploy the first smart contract on the blockchain, analyze the multi-dimensional data of each batch of cured pork products through the first smart contract, obtain the detection results of potential hazard factors, write the detection results of potential hazard factors into the blockchain and bind them to the batch information. S103. Add PUF labels to cured pork products and bind the PUF labels to batch information; S104. Identify the PUF label, query batch information based on the identification results, and further obtain the corresponding potential hazard factor detection results.

[0006] Furthermore, multi-dimensional data is collected during the production and storage of cured pork products, specifically including the following operations: S201. Establish an information archive for cured pork products, wherein the information archive shall include at least the origin, age, unique identification information, slaughter time, health status, slaughter time, slaughter location, and raw material test results of the source pigs; S202. At each key point in the production and storage of cured pork products, record the dwell time, environmental parameters, and processing techniques, and update them in the information archive. S203. At each key node in the production and storage process of cured pork products, relevant video images are recorded simultaneously. The video images are hashed to obtain hash values, and the video images are stored in a distributed storage system to obtain the storage path of the video images in the distributed storage system. S204. Write the hash value and storage path of the video image into the information file corresponding to the cured pork product.

[0007] Furthermore, a second smart contract is deployed in the blockchain. This second smart contract is configured to be triggered when mutually bound batch information, real-time status, and multi-dimensional data are uploaded to the blockchain. After being triggered, the second smart contract performs the following operations: S301. Extract the hash value and storage path from the multi-dimensional data, query the distributed storage system according to the storage path, and obtain the video image. S302. Perform a hash operation on the acquired video image to obtain a verification hash value. Compare the hash value extracted from the multi-dimensional data with the verification hash value to obtain the comparison result. S303. Input the video image into the image content recognition model to obtain the image content recognition result; S304. Vectorize the image content recognition results and their corresponding real-time status or multi-dimensional data respectively to obtain the result vector and the data vector, and calculate the similarity between the result vector and the data vector. S305. Based on the comparison results and similarity calculation results, determine whether to allow mutually bound batch information, real-time status, and multi-dimensional data to be uploaded to the blockchain.

[0008] Furthermore, the first smart contract is used to analyze multi-dimensional data of each batch of cured pork products to obtain the detection results of potential hazard factors, specifically including the following operations: S401. Based on the place of origin, slaughter time, slaughter time, and slaughter location in the information archive, query the historical breeding-related events that occurred at the place of origin and slaughter location at the corresponding time. S402. Input historical aquaculture-related events into the event analysis model for processing to obtain risk event analysis results; S403. Input the raw material test results, residence time at each key node, environmental parameters, and processing technology into the pathogen growth prediction model for processing to obtain pathogen growth prediction results. S404. Generate and output the detection results of potential hazard factors based on the risk event analysis results and pathogen growth prediction results.

[0009] Furthermore, after obtaining the detection results of potential hazard factors, perform the following operations: S501. Determine whether there are potential hazards based on the detection results. If so, proceed to the next step. S502. Extract entities from multi-dimensional data, aggregate information related to entities in multi-dimensional data through relational graph convolutional networks, analyze the relationships between entities, and construct a knowledge graph of potential hazards of cured pork products at the current moment. S503. Query the historical knowledge graph of potential hazards of cured pork products, aggregate the information of the historical knowledge graph of potential hazards of cured pork products through a multilayer perceptron, and use an activation function to obtain the historical prediction results of potential hazard risks. S504. Use a multilayer perceptron and a GRU network to aggregate the knowledge graph of potential hazards of cured pork products at the current moment, and use an activation function to obtain the prediction results of potential hazard risks at the current moment. S505. The historical prediction results and the current prediction results of potential hazards are weighted and summed to obtain the final prediction result, and the potential hazard risk level is determined based on the final prediction result.

[0010] Furthermore, the PUF tag includes a tag body, on which a unit array composed of a plurality of phase change material units is provided, a first power supply interface and a reset pulse generator, the first power supply interface being electrically connected to the phase change material units and the reset pulse generator respectively through a power supply circuit, and the reset pulse generator being electrically connected to the phase change material units.

[0011] Furthermore, the PUF tag is identified by a reading device, which includes a power supply, a second power supply interface, a current comparator, a communication module, a display screen, and a main control module. The power supply and the current comparator are electrically connected to the second power supply interface, and the main control module is connected to the current comparator and the display screen. The communication module is used to realize data interaction between the main control module and the host computer.

[0012] Furthermore, the PUF tag is identified, specifically including the following operations: S601. Connect the second power supply interface of the reading device to the first power supply interface of the PUF tag; S602. The readout device applies the same voltage to each phase change material unit to generate an actual current. S603. The reading device reads the actual current value generated by each phase change material unit through the current comparator and compares it with the reference current value to obtain the comparison result. S604. The main control module of the readout device constructs an identification matrix based on the comparison results corresponding to each phase change material unit in the unit array. S605 The main control module queries the detection results of potential hazard factors for the corresponding batch of cured pork products through the identification matrix and controls the display screen to show them.

[0013] Furthermore, after identifying the PUF tag using the reading device, the following operations are performed: S701. Activate the reset pulse generator. The reset pulse generator applies a reset electrical pulse to the random phase change material unit, causing the crystal state of the phase change material unit to change. S702. Repeat steps S602~S604 to obtain the reset recognition matrix; S703 The main control module uploads the reset identification matrix and the original identification matrix to the server through the communication module. After receiving the data, the server updates the identification matrix corresponding to the batch of cured pork products.

[0014] Compared with the prior art, the beneficial effects of the present invention are: This invention provides a method for detecting and tracing potential hazards in cured pork products. It collects multi-dimensional data from the production and storage stages of cured pork products, binds this data to batch information and real-time status, stores it using blockchain, and uses smart contracts to detect potential hazards. This shifts the focus from post-event traceability to pre-event warning. Through joint analysis of data from multiple stages, it significantly improves traceability efficiency and accuracy. Furthermore, based on PUF tags, it enables the querying of potential hazard detection results for different batches of cured pork products, solving the problem of easy counterfeiting of traditional static QR code traceability and further enhancing the anti-counterfeiting capabilities and credibility of the safety traceability of cured pork products. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only preferred embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of the overall process of a method for detecting and tracing potential hazards in cured pork products provided in an embodiment of the present invention.

[0017] Figure 2 This is a schematic diagram of the overall structure of the PUF label provided in an embodiment of the present invention.

[0018] Figure 3 This is a schematic diagram of the overall structure of the readout device provided in an embodiment of the present invention. Detailed Implementation

[0019] The principles and features of the present invention are described below with reference to the accompanying drawings. The listed embodiments are only used to explain the present invention and are not intended to limit the scope of the present invention.

[0020] Reference Figure 1 This embodiment provides a method for detecting and tracing potential hazard factors in cured pork products, the method comprising: S101. Collect multi-dimensional data during the production and storage of cured pork products, generate corresponding batch information, monitor the real-time status of cured pork products, and bind the real-time status, multi-dimensional data and batch information.

[0021] S102. Upload the mutually bound batch information, real-time status, and multi-dimensional data to the blockchain, deploy the first smart contract on the blockchain, analyze the multi-dimensional data of each batch of cured pork products through the first smart contract, obtain the detection results of potential hazard factors, write the detection results of potential hazard factors into the blockchain and bind them to the batch information.

[0022] For example, the blockchain is a consortium blockchain, and the nodes of the consortium blockchain include, but are not limited to, production enterprise nodes, regulatory department nodes, and user nodes.

[0023] S103. Add PUF labels to cured pork products and bind the PUF labels to batch information.

[0024] S104. Identify the PUF label, query batch information based on the identification results, and further obtain the corresponding potential hazard factor detection results.

[0025] The method provided in this embodiment, on the one hand, collects multi-dimensional data during the production and storage of cured pork products and detects their real-time status. The real-time status and multi-dimensional data are then bound to batch information and uploaded to the blockchain as the original data for traceability. Next, a smart contract deployed on the blockchain is used to detect potential hazard factors in different batches of cured pork products. The detection results are then written into the blockchain and bound to the batch information, thus ensuring that the original traceability data and potential hazard factor detection results are not tampered with, based on the inherent technical characteristics of the blockchain. On the other hand, by adding PUF tags to cured pork products and binding the PUF tags to batch information, users can obtain the potential hazard factor detection results for the corresponding batch of cured pork products by identifying the PUF tag. The PUF tag is unique and cannot be copied, which enhances the anti-counterfeiting capability and credibility of the traceability results.

[0026] As one possible implementation method, multi-dimensional data is collected during the production and storage of cured pork products, specifically including the following operations: S201. Establish an information archive for cured pork products, wherein the information archive shall include at least the origin, age, unique identification information, slaughter time, health status, slaughter time, slaughter location, and raw material test results of the source pig.

[0027] For example, the raw material testing results refer to the testing results of pork raw materials used to make cured meat products, including but not limited to all meat quality inspection contents stipulated in the meat quality inspection procedures for slaughtered pigs.

[0028] S202. At each key point in the production and storage of cured pork products, record the dwell time, environmental parameters, and processing techniques, and update them in the information archive.

[0029] S203. At each key node in the production and storage process of cured pork products, relevant video images are recorded simultaneously. The video images are hashed to obtain hash values, and the video images are stored in a distributed storage system to obtain the storage path of the video images in the distributed storage system.

[0030] S204. Write the hash value and storage path of the video image into the information file corresponding to the cured pork product.

[0031] In the production, processing, and storage of cured pork products, many factors can influence the formation of potential chemical hazards. For example, the long production cycle of traditional cured pork products, coupled with direct contact with oxygen and natural fermentation, leads to severe lipid oxidation, affecting their flavor and quality. The final products of lipid oxidation can induce various chronic diseases in humans and are one of the main factors contributing to the chemical hazards of traditional cured pork products. Curing is essential for the production of traditional cured pork products; to extend shelf life, large amounts of salt and nitrites are often used. Consequently, the residual salt and nitrite levels in traditionally cured meat products are generally high. High nitrite content can cause poisoning and easily combines with secondary amines to form nitrosamines, a class of carcinogenic compounds. The smoking process in the production of cured pork products can easily form carcinogenic polycyclic aromatic hydrocarbons (PAHs). Therefore, this implementation method first establishes an information archive based on static information such as the origin and unique identification information of the cured pork products. Then, at each key node in the production and storage process of cured pork products, dynamic data such as residence time, environmental parameters, and processing techniques are recorded and updated in the information archive. This implementation method preserves the static information of raw materials and the dynamic data of key production, processing, and storage stages of cured pork products through the information archive, so that the potential hazard factors of cured pork products can be analyzed based on these different dimensions of data. At the same time, relevant video footage is recorded as credible proof of the authenticity of the data for subsequent traceability verification.

[0032] As a further possible implementation, a second smart contract is deployed in the blockchain. This second smart contract is configured to be triggered when interlocked batch information, real-time status, and multi-dimensional data are uploaded to the blockchain. When the second smart contract is triggered, it performs the following operations: S301. Extract the hash value and storage path from the multi-dimensional data, query the distributed storage system based on the storage path, and obtain the video image.

[0033] S302. Perform a hash operation on the acquired video image to obtain a verification hash value. Compare the hash value extracted from the multi-dimensional data with the verification hash value to obtain the comparison result.

[0034] S303. Input the video image into the image content recognition model to obtain the image content recognition result.

[0035] In this embodiment, the image content recognition model is a neural network model that has been pre-trained using labeled sample video images and sample recognition results. The neural network model can be EfficientNet, SlowFast, or other neural networks that can achieve video content recognition. This embodiment does not specifically limit this.

[0036] S304. Vectorize the image content recognition results and their corresponding real-time status or multi-dimensional data to obtain result vectors and data vectors, and calculate the similarity between the result vectors and data vectors.

[0037] S305. Based on the comparison results and similarity calculation results, determine whether to allow mutually bound batch information, real-time status, and multi-dimensional data to be uploaded to the blockchain.

[0038] In this implementation, when uploading batch information, real-time status, and multi-dimensional data to the blockchain, the second smart contract first retrieves video footage, calculates a verification hash value, and compares it with the hash value extracted from the multi-dimensional data to obtain a comparison result. Then, it uses an image content recognition model to identify the video content and calculates the similarity between the vectorized image content recognition result and the real-time status or multi-dimensional data. Finally, based on the comparison result and similarity, it determines whether the real-time status corresponds to the multi-dimensional data. If they correspond, it means the video content can prove the authenticity of the real-time status or multi-dimensional data, and the mutually bound batch information, real-time status, and multi-dimensional data are allowed to be uploaded to the blockchain; otherwise, they are not allowed to be uploaded to the blockchain.

[0039] As another possible implementation, the first smart contract is used to analyze multi-dimensional data of each batch of cured pork products to obtain the detection results of potential hazard factors, specifically including the following operations: S401. Based on the place of origin, slaughter time, slaughter time, and slaughter location in the information archive, query the historical breeding-related events that occurred at the place of origin and slaughter location at the corresponding time.

[0040] For example, the historical aquaculture-related events include, but are not limited to: expansion / reduction of aquaculture sites, upgrading / sale of aquaculture equipment, livestock disease and death, and severe weather events in the aquaculture area.

[0041] S402. Input historical aquaculture-related events into the event analysis model for processing to obtain risk event analysis results.

[0042] In this embodiment, the event analysis model is a neural network model that has been pre-trained using historical aquaculture-related events and risk event analysis results from the samples.

[0043] S403. Input the raw material test results, residence time at each key node, environmental parameters, and processing technology into the pathogen growth prediction model for processing to obtain pathogen growth prediction results.

[0044] In this embodiment, the pathogen growth prediction model is a neural network model that has been pre-trained using sample dwell time, sample environmental parameters, sample processing technology, and sample pathogen growth prediction results.

[0045] S404. Generate and output the detection results of potential hazard factors based on the risk event analysis results and pathogen growth prediction results.

[0046] In this implementation, the detection of potential hazards in cured pork products includes two aspects: risk event analysis and pathogen growth prediction. Risk event analysis focuses on whether events that may occur during the breeding process of the raw materials for cured pork products could lead to the presence of pathogens in pigs. Pathogen growth prediction, based on the raw material detection results, considers the residence time, environmental parameters, and processing techniques of cured pork products at various key stages of production, processing, and storage. It predicts the growth of pathogens (such as Staphylococcus aureus, Salmonella, Listeria, pathogenic Escherichia coli, etc.), molds, parasites, etc., thereby identifying potential hazards in cured pork products and improving the efficiency and accuracy of the analysis.

[0047] As a further possible implementation, after obtaining the detection results of potential hazard factors, the following operations are performed: S501. Determine whether there are potential hazards based on the detection results. If so, proceed to the next step.

[0048] S502. Extract entities from multi-dimensional data, aggregate information related to entities in multi-dimensional data through a relational graph convolutional network, analyze the relationships between entities, and construct a knowledge graph of potential hazards of cured pork products at the current moment.

[0049] Understandably, each entity in multi-dimensional data has a unique identifier and associated attributes, such as name, type, description, and related entities. Relationships between entities are represented through relationships, which can be unidirectional or bidirectional.

[0050] S503. Query the historical knowledge graph of potential hazards of cured pork products, aggregate the information of the historical knowledge graph of potential hazards of cured pork products through a multilayer perceptron, and use an activation function to obtain the historical prediction results of potential hazard risks.

[0051] In this implementation, the current knowledge graph of potential hazards of cured pork products will be automatically transformed into a historical knowledge graph of potential hazards of cured pork products in the future.

[0052] S504. Use a multilayer perceptron and a GRU network to aggregate the knowledge graph of potential hazards of cured pork products at the current moment, and use an activation function to obtain the current moment prediction results of potential hazard risks.

[0053] S505. The historical prediction results and the current prediction results of potential hazards are weighted and summed to obtain the final prediction result, and the potential hazard risk level is determined based on the final prediction result.

[0054] This implementation method, upon detecting potential hazards in any batch of cured pork products, extracts entities and relationships from multi-dimensional data using a relational graph convolutional network to construct a knowledge graph of potential hazards for the cured pork products at the current moment. The relational graph convolutional network, through distributed computing and sparse matrix operations, processes the complex relationships between nodes, achieving efficient computation. It has advantages in handling coupled risk relationships in the production and storage of cured pork products and can also provide interpretations of the prediction results, helping users understand the model's predictions. Then, the knowledge graphs of potential hazards for cured pork products at the current moment and historical times are analyzed using a multilayer perceptron and a multilayer perceptron + GRU network, respectively, to obtain prediction results. This allows for learning from current and historical information to obtain the conditional distribution of potential hazard risk prediction. Finally, a weighted summation is used to obtain the final prediction result, determining the risk level corresponding to the potential hazard factor and helping users determine the risk of the potential hazard factor. This implementation method, by constructing a knowledge graph, better understands the relationships and interactions between various entities in the production and storage of cured pork products and predicts the risk level of potential hazard factors, improving the accuracy of safety risk prediction and early warning for cured pork products.

[0055] As another possible implementation method, refer to Figure 2 The PUF tag includes a tag body with a unit array composed of several phase change material units, a first power supply interface, and a reset pulse generator. The first power supply interface is electrically connected to the phase change material units and the reset pulse generator respectively through a power supply circuit, and the reset pulse generator is electrically connected to the phase change material units.

[0056] For example, the phase change material unit can be made of a phase change material, such as a germanium-antimony-tellurium phase change material. The phase change material can reversibly transition between a crystalline and amorphous state, and the PUF tag provided in this embodiment generates electrical pulses via a reset pulse generator to control this phase transition. In this embodiment, the shape and size of the amorphous region of each phase change material unit are random and sensitive to dynamic processes of heating and cooling, resulting in a continuously random distribution of resistance values.

[0057] The PUF tag provided in this embodiment is identified by a reading device. (See reference...) Figure 3 The reading device includes a power supply, a second power supply interface, a current comparator, a communication module, a display screen, and a main control module. The power supply and the current comparator are electrically connected to the second power supply interface, and the main control module is connected to the current comparator and the display screen. The communication module is used to realize data interaction between the main control module and the host computer.

[0058] Identifying PUF tags involves the following steps: S601. Connect the second power supply interface of the reading device to the first power supply interface of the PUF tag.

[0059] S602, The readout device applies the same voltage to each phase change material unit to generate actual current.

[0060] S603. The reading device reads the actual current value generated by each phase change material unit through the current comparator and compares it with the reference current value to obtain the comparison result.

[0061] S604. The main control module of the readout device constructs an identification matrix based on the comparison results of each phase change material unit in the unit array.

[0062] In this embodiment, each element in different rows and columns of the identification matrix corresponds to a phase change material unit in the unit array, and the element value is determined based on the comparison result. Since the resistance value of each phase change material unit is different, the corresponding actual current value also differs. The identification matrix constructed in this way serves as the identification information for the PUF tag and is bound to the detection results of potential hazard factors in different batches of cured pork products, thereby achieving a one-to-one mapping query. For example, the reference current value can be obtained by applying the same voltage to a reference resistor. When the actual current value is greater than the reference current value, the comparison result is assigned a value of 1; otherwise, it is assigned a value of 0. In some more specific embodiments, the actual current value can also be directly assigned to the elements in the identification matrix.

[0063] S605 The main control module queries the detection results of potential hazard factors for the corresponding batch of cured pork products through the identification matrix and controls the display screen to show them.

[0064] In this implementation, the identification matrix corresponding to the PUF tag is pre-stored in a server or blockchain. After the identification matrix corresponding to the PUF tag is identified by the reading device, the read identification matrix is ​​matched with the identification matrix pre-stored in the server or blockchain to obtain the detection results of potential hazard factors for the batch of cured pork products bound to the identification matrix.

[0065] Traditional traceability methods typically use QR codes or barcodes to query food traceability information, which are easily counterfeited, and the stored information can be easily copied or transferred. This implementation method solves the problem of counterfeitability of traditional traceability labels by using PUF tags with random microstructures, greatly enhancing the anti-counterfeiting capabilities and credibility of the traceability system.

[0066] Based on this, after the PUF tag is identified by the reading device, the following operations are performed: S701. Activate the reset pulse generator. The reset pulse generator applies a reset electrical pulse to the random phase change material unit, causing the crystal state of the phase change material unit to change.

[0067] S702. Repeat steps S602~S604 to obtain the reset recognition matrix.

[0068] S703 The main control module uploads the reset identification matrix and the original identification matrix to the server through the communication module. After receiving the data, the server updates the identification matrix corresponding to the batch of cured pork products.

[0069] For example, after receiving the reset recognition matrix and the recognition matrix before the reset, the server first queries the corresponding batch information based on the recognition matrix before the reset, and then updates the recognition matrix bound to the batch information based on the reset recognition matrix.

[0070] This implementation applies a reset electrical pulse to the random phase change material units in the PUF tag unit array via a reset pulse generator after each PUF tag identification by the reading device. This changes the crystal state of the random material units, thereby altering their resistance values ​​and resetting the identification matrix. This allows the static structure of the PUF tag to change dynamically after each identification, preventing attackers from simulating or cloning the corresponding information by establishing a digital model after obtaining the physical properties of the PUF tag. This further enhances the anti-counterfeiting capabilities and credibility of the traceability system.

[0071] 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 within the protection scope of the present invention.

Claims

1. A method for detecting potential hazard factors of cured pork products, characterized in that, The method comprises: S101, collecting multi-dimensional data in the production and storage process of cured pork products, generating corresponding batch information, monitoring the real-time state of the cured pork products, and binding the real-time state, multi-dimensional data and batch information; S102, uploading the mutually bound batch information, real-time state and multi-dimensional data to the blockchain, deploying a first smart contract on the blockchain, analyzing the multi-dimensional data of each batch of cured pork products through the first smart contract, obtaining potential hazard factor detection results, writing the potential hazard factor detection results into the blockchain and binding them with the batch information; S103, adding a PUF tag to the cured pork products and binding the PUF tag with the batch information; S104, identifying the PUF tag, querying the batch information according to the identification result and further obtaining the corresponding potential hazard factor detection result.

2. The method for detecting potential hazard factors of cured pork products according to claim 1, characterized in that, The multi-dimensional data is collected in the production and storage process of cured pork products, which specifically includes the following operations: S201, establishing an information file of cured pork products, which at least includes the origin, age, unique identification information, slaughter time, health status, slaughter time, slaughter site, raw material detection result of the source pig; S202, recording the residence time, environmental parameters and processing technology at each key node in the production and storage process of cured pork products, and updating them into the information file; S203, recording related video images at each key node in the production and storage process of cured pork products, performing hash operation on the video images to obtain hash values, storing the video images into a distributed storage system, and obtaining the storage path of the video images in the distributed storage system; S204, writing the hash value and storage path of the video images into the information file corresponding to the cured pork products.

3. The method for detecting potential hazard factors of cured pork products according to claim 2, characterized in that, The second smart contract is deployed in the blockchain, which is configured to be triggered when the mutually bound batch information, real-time state and multi-dimensional data are uploaded to the blockchain. After the second smart contract is triggered, the following operations are performed: S301, extracting the hash value and storage path in the multi-dimensional data, querying the distributed storage system according to the storage path, and obtaining the video image; S302, performing hash operation on the obtained video image to obtain a verification hash value, comparing the hash value extracted from the multi-dimensional data with the verification hash value to obtain a comparison result; S303, inputting the video image into an image content recognition model to obtain an image content recognition result; S304, vectorizing the image content recognition result and its corresponding real-time state or multi-dimensional data respectively to obtain a result vector and a data vector, and calculating the similarity of the result vector and the data vector; S305, determining whether to allow the mutually bound batch information, real-time state and multi-dimensional data to be uploaded to the blockchain according to the comparison result and the similarity calculation result.

4. The method for detecting potential hazard factors of cured pork products according to claim 2, characterized in that, The first smart contract is used to analyze the multi-dimensional data of each batch of cured pork products to obtain potential hazard factor detection results, which specifically includes the following operations: S401, according to the origin information, the time of delivery, the time of slaughter, the place of slaughter in the information file, the historical breeding related events occurred in the corresponding time and place are inquired; S402, the historical breeding related events are input into the event analysis model for processing to obtain the risk event analysis result; S403, the raw material detection result, the residence time of each key node, the environmental parameters and the processing technology are input into the bacteria growth prediction model for processing to obtain the bacteria growth prediction result; S404, the potential hazard factor detection result is generated according to the risk event analysis result and the bacteria growth prediction result and is output.

5. The method for detecting potential hazard factors of cured pork products according to claim 4, characterized in that, After obtaining the potential hazard factor detection result, the following operations are performed: S501, whether there is a potential hazard factor is judged according to the potential hazard factor detection result, if yes, the next step is performed; S502, entities are extracted from multidimensional data, information related to the entities in the multidimensional data is aggregated through a relation graph convolution network, the relationship between the entities is analyzed, and a potential hazard knowledge graph of cured pork products at the current time is constructed; S503, the historical potential hazard knowledge graph of cured pork products is inquired, the information of the historical potential hazard knowledge graph of cured pork products is aggregated through a multilayer perceptron, and a historical prediction result of the potential hazard risk is obtained using an activation function; S504, the potential hazard knowledge graph of cured pork products at the current time is aggregated using a multilayer perceptron and a GRU network, and a current time prediction result of the potential hazard risk is obtained using an activation function; S505, the historical prediction result and the current time prediction result of the potential hazard risk are weighted and summed to obtain a final prediction result, and the potential hazard risk level is determined according to the final prediction result.

6. The method for detecting potential hazard factors of cured pork products according to claim 1, characterized in that, The PUF tag includes a tag body, a unit array composed of a plurality of phase change material units is arranged on the tag body, a first power supply interface and a reset pulse generator, the first power supply interface is electrically connected with the phase change material units and the reset pulse generator through a power supply circuit, and the reset pulse generator is electrically connected with the phase change material units.

7. The method for detecting potential hazard factors of cured pork products according to claim 6, characterized in that, The PUF tag is identified by a readout device, the readout device includes a power supply, a second power supply interface, a current comparator, a communication module, a display screen and a main control module, the power supply and the current comparator are electrically connected with the second power supply interface, the main control module is signal connected with the current comparator and the display screen, and the communication module is used for realizing data interaction between the main control module and an upper computer.

8. The method according to claim 7, wherein the method is for detecting potential hazard factors of a cured pork product, characterized in that, The PUF tag is identified, and the specific operations include the following steps: S601, the second power supply interface of the readout device is connected with the first power supply interface of the PUF tag; S602, the readout device applies the same voltage to each phase change material unit to generate an actual current; S603, the readout device reads the actual current value generated by each phase change material unit through the current comparator and compares it with a reference current value to obtain a comparison result; S604, the main control module of the readout device constructs an identification matrix according to the comparison results corresponding to each phase change material unit in the unit array. S605, the master module queries the potential hazard factor detection result of the corresponding batch of cured pork products through the identification matrix and controls the display screen to display.

9. The method for detecting potential hazard factors of cured pork products according to claim 8, characterized in that, After the PUF tag is identified by the reading device, the following operations are performed: S701, activate the reset pulse generator, the reset pulse generator applies a reset electric pulse to the random phase change material unit, so that the crystal state of the phase change material unit changes; S702, re-execute steps S602-S604 to obtain a reset identification matrix; S703, the master module uploads the reset identification matrix and the identification matrix before the reset to the server through the communication module, and the server receives and updates the corresponding identification matrix of the corresponding batch of cured pork products.

Images