Fresh agricultural product full-link traceability management system

Abstract

The invention relates to the technical field of data processing, and discloses a fresh agricultural product full-link traceability management system, which comprises a data acquisition module used for acquiring original data representing the state of the data acquisition module; the artificial intelligence analysis module is used for processing the original data to generate an analysis result and a credibility score corresponding to the analysis result; an intelligent contract is deployed in the data write-in scheduling module, and the intelligent contract is used for receiving the analysis result and the credibility score; and the hierarchical block chain storage module is used for responding to the write-in instruction generated by the data write-in scheduling module. Through intelligent analysis and hierarchical block chain storage, quantitative evaluation and dynamic hierarchical storage management of source credibility of fresh agricultural product traceability data are realized, a main chain is effectively prevented from being polluted by abnormal data, a data closed-loop recheck mechanism is established, and the overall authenticity and reliability of traceability information are improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a full-chain traceability management system for fresh agricultural products. Background Technology

[0002] In the supply chain management of fresh agricultural products, establishing a complete and reliable product traceability system is crucial. Consumers are increasingly demanding higher standards of food safety and quality transparency. A system that clearly displays information about the entire process of a product from its origin to its table is not only fundamental to protecting public health but also a core element for businesses to build brand trust and enhance market competitiveness.

[0003] To meet these needs, existing technologies widely employ traceability solutions that combine IoT devices with blockchain technology. In this solution, sensors deployed in production, warehousing, and logistics automatically collect environmental and product data and directly submit this data to a unified blockchain ledger. Leveraging the decentralized, consensus-driven, and chain-like data structure of blockchain, this solution ensures that once data is recorded, its content and timestamp cannot be tampered with. This technology provides a shared, immutable trust ledger for multiple participants in the supply chain, effectively solving the problems of data susceptibility to tampering and difficulty in establishing trust in traditional centralized databases.

[0004] However, while the aforementioned existing technologies ensure the "immutability" of data after it is uploaded to the blockchain, they fail to address the issue of "source credibility" before data is uploaded. This technical solution treats all raw data collected from IoT devices equally, lacking a pre-emptive, automated data quality screening and credibility assessment mechanism. This results in abnormal or low-quality data caused by sensor malfunctions, environmental interference, or human error being indiscriminately written into the blockchain, causing permanent information pollution to the main traceability records. Furthermore, its single blockchain storage architecture lacks flexibility, failing to differentiate and isolate data of different credibility levels. This not only reduces the purity and query efficiency of core traceability data but also lacks a closed-loop management system for subsequent verification and status confirmation of questionable data, ultimately affecting the overall reliability of the entire traceability system. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a full-chain traceability management system for fresh agricultural products. It solves the problem that existing technologies directly upload raw data collected by IoT devices to the blockchain, lacking a mechanism for credibility assessment and quality verification of the data source. This results in questionable or erroneous traceability data being directly written to the blockchain, which not only contaminates the core traceability records, but also fails to achieve effective isolation and closed-loop management of such data in the existing single storage architecture.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A first aspect of the present invention provides a full-chain traceability management system for fresh agricultural products. The system includes:

[0008] The data acquisition module is used to collect raw data that characterizes the state of fresh agricultural products during the traceability process.

[0009] An artificial intelligence analysis module, connected to the data acquisition module, is used to receive the raw data and process the raw data to generate analysis results and a corresponding credibility score.

[0010] The data writing scheduling module is connected to the artificial intelligence analysis module. The data writing scheduling module is equipped with a smart contract. The smart contract is used to receive the analysis results and the credibility score, and generate writing instructions for the main chain or the side chain to be verified according to the preset high credibility threshold and low credibility threshold.

[0011] The hierarchical blockchain storage module, including the main chain and the side chain to be verified, is used to respond to the write command generated by the data write scheduling module and selectively store the analysis results in the main chain or the side chain to be verified.

[0012] Preferably, the data acquisition module includes:

[0013] An environmental status acquisition unit is used to collect environmental parameters of the source tracing process as part of the raw data.

[0014] A visual information acquisition unit is used to acquire image information of the fresh agricultural products as another part of the raw data.

[0015] In one specific embodiment, the artificial intelligence analysis module includes:

[0016] The data preprocessing unit is used to format and filter noise from the raw data; the feature extraction unit is used to extract feature vectors for analysis from the preprocessed raw data.

[0017] The analysis model execution unit is used to run a preset analysis model and generate the analysis results and the model's own confidence level based on the feature vector.

[0018] The credibility assessment unit is used to calculate the final credibility score by combining the model's own confidence level, the data quality factor calculated based on the original data, and the cross-validation consistency weight.

[0019] Preferably, the credibility assessment unit is specifically used to calculate the data quality factor based on the clarity of the image information or the signal-to-noise ratio of the environmental parameters in the original data, and to compare the analysis results generated by the analysis model execution unit with preset agricultural activity logs or historical traceability data to determine the cross-validation consistency weight.

[0020] Furthermore, the data writing scheduling module includes:

[0021] A threshold judgment unit is used to compare the received confidence score with the high confidence threshold and the low confidence threshold to generate a judgment result;

[0022] The instruction generation unit is used to generate, based on the judgment result, the write instruction to write the analysis result into the main chain or the side chain to be verified.

[0023] Preferably, the instruction generation unit is specifically used to generate a status label in the write instruction when the judgment result is to write the analysis result into the side chain to be verified; the status label is used to characterize the pending status of the analysis result, and the pending status includes a pending manual review status or a high-risk conflict alarm status.

[0024] In one specific embodiment, the hierarchical blockchain storage module includes:

[0025] The main chain storage unit is used to store the analysis results that the data writing scheduling module determines should be written to the main chain, so as to form an immutable traceability record;

[0026] A sidechain storage unit is used to store the analysis results that the data writing scheduling module determines should be written to the sidechain to be verified, so as to isolate the traceability information to be verified.

[0027] The dynamic verification processing unit is used to update or anchor the status of the analysis results in the main chain storage unit after receiving an external audit confirmation instruction for the analysis results stored in the side chain storage unit.

[0028] Preferably, the dynamic verification processing unit is specifically used to: receive the external audit confirmation instruction, wherein the external audit confirmation instruction includes authorization verification information and an identifier corresponding to a certain analysis result in the sidechain storage unit;

[0029] The authorization verification information is verified to confirm the legality of the external audit confirmation instruction; after the authorization verification information passes the verification, a verification transaction is generated, which includes the identifier and the confirmed status.

[0030] The verification transaction is submitted to the main chain storage unit so that the traceability status of the analysis result corresponding to the identifier is updated to confirmed.

[0031] A second aspect of the present invention provides a blockchain-based intelligent traceability management method for enterprise supply chains, which is executed by the system described in any of the foregoing embodiments, characterized by comprising the following steps:

[0032] S1. Using the data acquisition module, collect raw data that characterizes the state of fresh agricultural products in the traceability process.

[0033] S2. Using an artificial intelligence analysis module, receive and process the raw data to generate analysis results and a credibility score corresponding to the analysis results;

[0034] S3. Using the data writing scheduling module, receive the analysis results and the credibility score, and compare the credibility score with the preset high credibility threshold and low credibility threshold through the smart contract to generate a write instruction for the main chain or the side chain to be verified.

[0035] S4. Using the hierarchical blockchain storage module, in response to the generated write command, the analysis results are selectively stored in the main chain or the side chain to be verified.

[0036] S5. When an external audit confirmation instruction is received for the analysis results stored in the sidechain to be verified, the dynamic verification processing unit in the layered blockchain storage module is used to update or anchor the status of the analysis results in the main chain.

[0037] Preferably, in step S3, generating write instructions for the main chain or the side chain to be verified specifically includes:

[0038] When the credibility score is not lower than the high credibility threshold, a write instruction is generated to write the analysis results to the main chain;

[0039] When the confidence score is lower than the high confidence threshold but not lower than the low confidence threshold, a write instruction is generated to write the analysis result into the sidechain to be verified.

[0040] When the credibility score is lower than the low credibility threshold, a write instruction is generated to write the analysis results into the sidechain to be verified.

[0041] This invention provides a full-chain traceability management system for fresh agricultural products. It has the following beneficial effects:

[0042] 1. This invention establishes an artificial intelligence analysis module to generate analysis results and corresponding credibility scores after data collection. The smart contract in the data writing scheduling module then judges these scores based on preset thresholds. This setup ensures that only analysis results meeting specific credibility requirements are accepted, thus establishing a pre-screening mechanism before data is written to the blockchain storage. This avoids interference from low-quality or questionable data on the traceability main chain, improving the overall reliability of the final traceability data.

[0043] 2. This invention employs a hierarchical blockchain storage module, comprising a main chain and a side chain to be verified. A data writing scheduling module selectively stores analysis results on different chains based on trust scores. This hierarchical storage design enables differentiated processing of data with different trust levels. It allows for the rapid storage of highly trustworthy data on the main chain while isolating medium- and low-trust data on the side chain to be verified, ensuring the purity of the main chain data and efficient querying.

[0044] 3. This invention, by setting up a dynamic verification processing unit in the layered blockchain storage module, allows the system to receive external audit confirmation instructions for data stored in the sidechain to be verified, and updates or anchors its status on the main chain after verification. This setup constructs a management closed loop from "isolation and doubt" to "audit and confirmation," enabling isolated data to have its trust level improved through manual review and other methods. While ensuring data integrity, it achieves trusted management of traceability information throughout its entire lifecycle. Attached Figure Description

[0045] Figure 1 This is a system architecture diagram of the present invention;

[0046] Figure 2 This is a schematic diagram of the method flow of the present invention.

[0047] Among them, 10 is the data acquisition module; 20 is the artificial intelligence analysis module; 30 is the data writing and scheduling module; and 40 is the hierarchical blockchain storage module. Detailed Implementation

[0048] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] Reference Figure 1 , Figure 1 This is a structural architecture diagram of a fresh agricultural product end-to-end traceability management system according to an embodiment of the present invention.

[0050] This invention provides a full-chain traceability management system for fresh agricultural products. The system includes: a data acquisition module 10, an artificial intelligence analysis module 20, a data writing and scheduling module 30, and a hierarchical blockchain storage module 40.

[0051] The data acquisition module 10 is connected to the artificial intelligence analysis module 20 and is used to send the raw data collected in the traceability process to the artificial intelligence analysis module 20.

[0052] The artificial intelligence analysis module 20 is connected to the data writing scheduling module 30 and is used to send the analysis results and credibility scores it generates to the data writing scheduling module 30.

[0053] The data write scheduling module 30 is connected to the hierarchical blockchain storage module 40 and is used to send write instructions to the hierarchical blockchain storage module 40.

[0054] The layered blockchain storage module 40 includes a main chain and a side chain to be verified, which are used to respond to the instructions of the data writing scheduling module 30.

[0055] Specifically, the data acquisition module 10 is used to collect raw data that characterizes the status of fresh agricultural products in the traceability process (such as planting, processing, warehousing, and logistics).

[0056] In one embodiment, the data acquisition module 10 includes an environmental status acquisition unit, used to acquire environmental parameters (such as temperature, humidity, soil composition, and light intensity) of the tracing process as part of the raw data;

[0057] The visual information acquisition unit is used to acquire image information of the fresh agricultural products (such as crop growth images and product sorting images) as another part of the raw data.

[0058] The artificial intelligence analysis module 20 receives the raw data from the data acquisition module 10. The AI ​​analysis module 20 processes the raw data to generate analysis results and a corresponding credibility score. In one embodiment, the AI ​​analysis module 20 includes: a data preprocessing unit, a feature extraction unit, an analysis model execution unit, and a credibility evaluation unit.

[0059] The data preprocessing unit formats, timestamps, and filters noise from the raw data. The feature extraction unit extracts feature vectors for analysis from the preprocessed raw data. The analysis model execution unit runs a pre-defined analysis model (e.g., a convolutional neural network model or a time series analysis model) and generates the analysis results and the model's own confidence score based on the feature vectors. The credibility evaluation unit combines the model's own confidence score, the data quality factor calculated from the raw data, and the cross-validation consistency weights to calculate the final credibility score.

[0060] A smart contract is deployed within the data write scheduling module 30. This module receives the analysis results and the credibility score from the artificial intelligence analysis module 20. The smart contract determines the received credibility score based on preset high and low credibility thresholds, and generates write instructions for the main chain or the side chain to be verified accordingly.

[0061] In one embodiment, the data writing scheduling module 30 includes: a threshold judgment unit, used to compare the received credibility score with the high credibility threshold and the low credibility threshold to generate a judgment result;

[0062] The instruction generation unit is used to generate, based on the judgment result, the write instruction to write the analysis result into the main chain or the side chain to be verified.

[0063] In a further embodiment, when the write instruction generated by the instruction generation unit points to the sidechain to be verified, the write instruction further includes a status tag. The status tag is used to characterize the pending status of the analysis result, which includes a pending manual review status or a high-risk conflict alarm status.

[0064] The hierarchical blockchain storage module 40 is used to respond to the write command from the data write scheduling module 30. The hierarchical blockchain storage module 40 includes a main chain storage unit (corresponding to the main chain) and a side chain storage unit (corresponding to the side chain to be verified). The main chain storage unit stores the analysis results determined by the data write scheduling module 30 to be written to the main chain, forming an immutable traceability record. The side chain storage unit stores the analysis results determined by the data write scheduling module 30 to be written to the side chain to be verified, isolating the traceability information to be verified.

[0065] In one embodiment, the hierarchical blockchain storage module 40 further includes a dynamic verification processing unit. This dynamic verification processing unit is used to update or anchor the status of the analysis results stored in the main chain storage unit after receiving an external audit confirmation instruction for the analysis results stored in the sidechain storage unit. The external audit confirmation instruction can be issued by a management terminal with preset permissions, and the instruction includes authorization verification information for verifying permissions.

[0066] Reference Figure 1 The data acquisition module 10 is the data source of the system of the present invention, used to automatically or semi-automatically collect raw data characterizing the status of fresh agricultural products (e.g., crops, livestock, or aquatic products) in multiple traceability stages. These traceability stages include, but are not limited to, production stages (e.g., planting, breeding), processing stages (e.g., sorting, packaging), warehousing, and logistics. The data acquisition module 10 specifically includes:

[0067] Environmental status acquisition unit. In one embodiment, the environmental status acquisition unit is an Internet of Things (IoT) sensor network, which is configured with different types of sensors depending on the traceability link being deployed.

[0068] In the production process (such as field planting or greenhouse), the environmental status acquisition unit includes: one or more soil sensors deployed in the soil for collecting soil temperature, humidity, electrical conductivity and pH value; and one or more micro weather stations deployed in the environment for collecting air temperature, humidity, light intensity and carbon dioxide concentration.

[0069] In the warehousing or logistics process, the environmental status acquisition unit includes: a high-precision temperature and humidity sensor deployed inside a cold chain vehicle or warehouse; a global positioning system module or a Beidou positioning module for acquiring geographical location information; and a vibration sensor or an unpacking sensor for monitoring the status of goods.

[0070] The environmental status acquisition unit acquires relevant environmental parameters at a preset acquisition cycle (e.g., every 5 minutes) or in response to a specific event (e.g., an unpacking event), and sends these environmental parameters as part of the raw data to the artificial intelligence analysis module 20.

[0071] The data acquisition module 10 further includes a visual information acquisition unit. The visual information acquisition unit is used to acquire image information of the fresh agricultural products themselves or image information of their growing environment.

[0072] In the production process, the visual information acquisition unit includes: a high-definition monitoring camera deployed at a fixed location for periodically capturing panoramic growth images of the crop; and a high-resolution visible light camera or multispectral camera integrated on a drone for patrolling and acquiring close-up images of the crop. The data acquired by the multispectral camera can be used to subsequently generate vegetation indices for the artificial intelligence analysis module 20 to assess the crop's growth and stress status.

[0073] In the processing stage (such as a sorting line), the visual information acquisition unit includes an industrial camera or line scan camera deployed above the production line, and a dedicated, uniform illumination source (e.g., a ring light source or a strip light source) to capture images of products as they pass by. The image information is used for subsequent analysis of the product's shape, color, maturity, and surface defects.

[0074] After acquiring the image information, the visual information acquisition unit sends it as another part of the raw data to the artificial intelligence analysis module 20.

[0075] In a preferred embodiment, the data acquisition module 10 further includes a data synchronization and association unit. This unit has a built-in unified clock synchronization mechanism to ensure that the raw data acquired by the environmental status acquisition unit and the visual information acquisition unit are all stamped with high-precision and consistent timestamps. Furthermore, the data synchronization and association unit is also used to logically bind environmental parameters and image information from the same traceability object or the same batch, acquired at similar time points (e.g., assigning them the same batch identifier), forming a raw data packet containing multimodal data, and then sending this data packet as a whole to the artificial intelligence analysis module 20 for subsequent cross-validation analysis.

[0076] Reference Figure 1 The artificial intelligence analysis module 20 is connected to the data output terminal of the data acquisition module 10 and is used to receive the raw data. The function of the artificial intelligence analysis module 20 is to perform in-depth processing and analysis on the raw data to generate structured analysis results and a credibility score that quantifies the reliability of the analysis results. Specifically, module 20 includes:

[0077] The data preprocessing unit performs preliminary processing on the received raw data. When environmental parameters (e.g., temperature and humidity sequences) are received, the data preprocessing unit performs numerical normalization on them (e.g., min-max normalization) to map data of different dimensions into intervals, thereby eliminating the impact of dimensional differences on subsequent model analysis. When image information is received, the data preprocessing unit performs size normalization on the image information (e.g., uniformly scaling it to 224x224 pixels) to match the input requirements of the analysis model, and can perform operations such as Gaussian filtering to filter out image noise.

[0078] The feature extraction unit is connected to the data preprocessing unit. This unit extracts feature vectors for analysis from the preprocessed raw data. For image information, when the subsequent analysis model execution unit uses a convolutional neural network, the feature extraction function is automatically completed by the convolutional and pooling layers of the convolutional neural network, extracting hierarchical features from edges and textures to local shapes through multiple convolutional operations. For time-series data such as environmental parameters, the feature extraction unit can extract its statistical features, such as mean, variance, kurtosis, and Fourier transform coefficients of the time series, to form feature vectors.

[0079] The analysis model execution unit is used to run preset analysis models. These models are configured according to the different traceability stages and data types. For example, in the production stage, a convolutional neural network model based on the ResNet architecture can be configured to analyze image information of crop leaves to identify pest and disease types. In the logistics stage, a model based on a Long Short-Term Memory (LSTM) network can be configured to analyze temperature sequence data during cold chain transportation to determine whether there are temperature anomalies or chain breakage risks.

[0080] The analysis model execution unit receives feature vectors from the feature extraction unit and outputs two items: the first is the analysis result R, which is the model's final judgment (e.g., "downy mildew" or "normal temperature"); the second is the model's own confidence level. This value is usually taken from the output probability of the Softmax layer at the end of the classification model, and represents the confidence level of the model in the judgment conclusion it has made.

[0081] The confidence assessment unit does not directly adopt the model's own confidence level. Instead of using a single, definitive credibility score, the final credibility score is calculated by integrating information from multiple dimensions. In one specific embodiment, the credibility score The following formula is used to calculate:

[0082]

[0083] In the formula, Assess the credibility of the final output; The confidence level of the model itself output by the analysis model execution unit; This is a data quality factor calculated based on the raw data; the calculation method of this factor is related to the data type. For image information, the data quality factor... Sharpness can be calculated based on the image's sharpness. For example, sharpness can be quantified by calculating the variance of the image's Laplacian operator response; the larger the variance, the sharper the image. The higher the value, the better. For sensor data such as environmental parameters, It can be calculated based on the signal-to-noise ratio (SNR). The higher the SNR, the better. The higher the value.

[0084] For cross-validation consistency weights; , , These are configurable weighting coefficients, used to adjust the influence of model confidence, data quality factor, and cross-validation consistency weight in the final score calculation.

[0085] After completing the calculation, the credibility assessment unit will output the final analysis result R and credibility score. As a data pair, it is sent to the data writing scheduling module 30.

[0086] Reference Figure 1 The data writing scheduling module 30 is a key component connecting the artificial intelligence analysis module 20 and the hierarchical blockchain storage module 40. Its core function is to receive analysis results and credibility scores, and, based on a set of deterministic rules, decide which layer of the blockchain the data should be written to. The logic of the data writing scheduling module 30 is implemented by smart contracts deployed in the hierarchical blockchain storage module 40 network, thereby ensuring the automation, transparency, and immutability of the scheduling process.

[0087] The data writing scheduling module 30 specifically includes:

[0088] The threshold judgment unit is used to receive the credibility score from the artificial intelligence analysis module 20. The threshold judgment unit has two preset configurable numerical thresholds: a high confidence threshold and a low confidence threshold. and low confidence threshold ,in The function of the threshold judgment unit is to determine the confidence score received. The result is compared with these two thresholds to generate a judgment. This judgment logic can be formally described by the following conditional expression:

[0089]

[0090] In the formula, The judgment result output by the threshold judgment unit; Assess credibility. The high confidence threshold; The threshold for low confidence level; Indicates "if"; This is one type of judgment result, indicating "high confidence"; This is one possible outcome, indicating "pending review". This is one type of judgment result, indicating "alarm".

[0091] The instruction generation unit, connected to the threshold determination unit, is used to determine the threshold value based on the received determination result. This generates a structured write instruction. This write instruction defines the target blockchain layer for the data write and the accompanying metadata. Specifically:

[0092] When the received judgment result is " When the instruction generation unit generates a write instruction pointing to the main chain, the write content contained in this instruction is the analysis result R generated by the artificial intelligence analysis module 20.

[0093] When the received judgment result is " When the command is executed, the instruction generation unit generates a write instruction pointing to the sidechain to be verified. In addition to the analysis result R, the instruction also includes a status label, the content of which is set to "awaiting manual review".

[0094] When the received judgment result is " At the same time, the instruction generation unit also generates a write instruction pointing to the sidechain to be verified. In addition to the analysis result R, the instruction also includes a status tag set to "high-risk conflict alarm status". This status tag can also be used to trigger an off-chain notification event, sending the alarm information to a preset management terminal via a message push service.

[0095] After the generation is completed, the instruction generation unit sends the structured write instruction to the hierarchical blockchain storage module 40, which then performs the subsequent blockchain write operation.

[0096] Reference Figure 1 The layered blockchain storage module 40 is the core of the system for data persistence and trust solidification in this invention. It receives write instructions from the data write scheduling module 30 and is responsible for storing data securely and traceably. This layered blockchain storage module 40 is physically or logically constructed as a multi-chain network, specifically including:

[0097] A main chain storage unit, corresponding to the main chain, is configured as a primary repository for highly reliable data. It is specifically designed to respond to data generated by the data write scheduling module 30 with a judgment result of "". The main chain storage unit receives a write instruction. When such an instruction is received, the main chain storage unit packages the analysis result R contained in the instruction into a transaction and writes it into the main chain block, thereby forming an immutable and highly reliable traceability record.

[0098] Sidechain storage unit 40b corresponds to the sidechain to be verified. This sidechain storage unit is configured as a data isolation zone for temporary or permanent storage of data with insufficient trust levels. It is specifically designed to respond to data whose judgment result is "...". "or" The sidechain storage unit receives a write instruction. When such an instruction is received, the sidechain storage unit stores the analysis result R contained in the instruction along with the corresponding status label (e.g., "awaiting manual review" or "high-risk conflict alarm").

[0099] The hierarchical blockchain storage module 40 also includes a dynamic verification processing unit. This dynamic verification processing unit establishes a status confirmation channel from the sidechain storage unit to the mainchain storage unit, used to achieve "trust upgrade" of data. The workflow of the dynamic verification processing unit is as follows:

[0100] The dynamic verification processing unit is configured to listen for or receive external audit confirmation instructions from an external entity with preset permissions (e.g., an administrator's operating terminal). In one embodiment, the external audit confirmation instruction is a digitally signed data structure, the contents of which specifically include: an identifier. This identifier uniquely points to a specific analysis result awaiting review in the sidechain storage unit; and includes authorization verification information. For example, an administrator uses their private key to generate a digital signature for the instruction (or its digest).

[0101] Upon receiving the external audit confirmation instruction, the dynamic verification processing unit executes a verification step. It utilizes the public key corresponding to the administrator, either pre-stored or retrieved from the access control contract. For the authorization verification information Decryption and comparison are performed to verify the legality and integrity of the instruction.

[0102] The dynamic verification processing unit is triggered to generate a new verification transaction only if the authorization verification information passes the verification. This verifies the transaction. Constructed to include the following information: the identifier and the confirmed status (For example, a boolean value "True" or an enumeration value "Verified").

[0103] Finally, the dynamic verification processing unit verifies the transaction. Submitted to the main chain storage unit. This verifies the transaction. Once packaged and written into the main chain, an immutable record is created on the main chain, which publicly and traceably confirms the identifier. The corresponding sidechain data has passed external review. This process updates or anchors the status of the analysis results within the main chain storage unit. Subsequent traceability queries can then determine the final confirmation status of this sidechain data when querying the main chain.

[0104] Reference Figure 2 , Figure 2 This is a flowchart of a method for full-chain traceability management of fresh agricultural products according to one embodiment of the present invention. Another embodiment of the present invention provides a method for full-chain traceability management of fresh agricultural products, which can be executed by the system in the foregoing embodiment. The method specifically includes the following steps:

[0105] Step S100: Using the data acquisition module 10, raw data characterizing the state of fresh agricultural products is collected in the traceability process. In a specific execution process, step S100 includes: collecting environmental parameters (e.g., temperature, humidity, light intensity) of the traceability process using the environmental state acquisition unit; and collecting image information (e.g., crop leaf images, product sorting images) of the fresh agricultural products using the visual information acquisition unit. After collection, the raw data containing timestamps and batch identifiers is sent to the artificial intelligence analysis module 20.

[0106] Step S200: The artificial intelligence analysis module 20 receives and processes the raw data. Step S200 first formats and filters noise from the raw data using a data preprocessing unit; then, a feature extraction unit extracts feature vectors from the preprocessed data; the analysis model execution unit runs a preset analysis model and generates analysis results and model confidence based on the feature vectors; finally, the credibility assessment unit combines the model confidence, the data quality factor calculated from the raw data, and the cross-validation consistency weights to calculate the final credibility score.

[0107] Step S300: The data writing scheduling module 30 receives the analysis results and the credibility score generated in step S200. The smart contract deployed within the data writing scheduling module 30 is then triggered to execute. Specifically, the threshold judgment unit compares the credibility score with preset high credibility thresholds and low credibility thresholds.

[0108] The instruction generation unit generates write instructions for the main chain or the sidechain to be verified based on the comparison and judgment result of the threshold judgment unit. This generation process specifically includes:

[0109] When the credibility score is not lower than the high credibility threshold, a write instruction is generated to write the analysis results to the main chain;

[0110] When the credibility score is lower than the high credibility threshold but not lower than the low credibility threshold, a write instruction is generated to write the analysis result to the sidechain to be verified. The instruction also includes a status label of "awaiting manual review".

[0111] When the credibility score is lower than the low credibility threshold, a write instruction is generated to write the analysis results to the sidechain to be verified, and a status label of "high-risk conflict alarm status" is attached.

[0112] Step S400: Utilize the hierarchical blockchain storage module 40 to respond to the write instruction generated in step S300. If the write instruction points to the main chain, the main chain storage unit performs an operation to store the analysis result in the main chain block. If the write instruction points to the sidechain to be verified, the sidechain storage unit performs an operation to store the analysis result and the corresponding status tag together in the sidechain to be verified, thereby achieving hierarchical and isolated storage of data with different trust levels.

[0113] Step S500: Regarding the analysis results stored in the sidechain to be verified in step S400, this method further includes a dynamic verification step. When the system receives an external audit confirmation instruction for the analysis results, the dynamic verification processing unit in the hierarchical blockchain storage module 40 is activated. The dynamic verification processing unit first verifies the authorization verification information (e.g., the administrator's digital signature) in the external audit confirmation instruction. After the verification passes, the dynamic verification processing unit generates a verification transaction containing the identifier of the analysis results in the sidechain and a "confirmed status," and submits the verification transaction to the main chain storage unit. After the verification transaction is written to the main chain, the update or anchoring of the status of the sidechain analysis results in the main chain is completed, forming a management closed loop.

Claims

1. A full-chain traceability management system for fresh agricultural products, characterized in that: include: The data acquisition module is used to collect raw data that characterizes the state of fresh agricultural products during the traceability process. An artificial intelligence analysis module, connected to the data acquisition module, is used to receive the raw data and process the raw data to generate analysis results and a corresponding credibility score. The data writing scheduling module is connected to the artificial intelligence analysis module. The data writing scheduling module is equipped with a smart contract. The smart contract is used to receive the analysis results and the credibility score, and generate writing instructions for the main chain or the side chain to be verified according to the preset high credibility threshold and low credibility threshold. The hierarchical blockchain storage module, including the main chain and the side chain to be verified, is used to respond to the write command generated by the data write scheduling module and selectively store the analysis results in the main chain or the side chain to be verified.

2. The fresh agricultural product end-to-end traceability management system according to claim 1, characterized in that, The data acquisition module includes: An environmental status acquisition unit is used to collect environmental parameters of the source tracing process as part of the raw data. A visual information acquisition unit is used to acquire image information of the fresh agricultural products as another part of the raw data.

3. The fresh agricultural product end-to-end traceability management system according to claim 1, characterized in that, The artificial intelligence analysis module includes: A data preprocessing unit is used to format and filter noise from the raw data; A feature extraction unit is used to extract feature vectors for analysis from the preprocessed raw data; The analysis model execution unit is used to run a preset analysis model and generate the analysis results and the model's own confidence level based on the feature vector. The credibility assessment unit is used to calculate the final credibility score by combining the model's own confidence level, the data quality factor calculated based on the original data, and the cross-validation consistency weight.

4. The fresh agricultural product end-to-end traceability management system according to claim 3, characterized in that, The credibility assessment unit is specifically used to calculate the data quality factor based on the clarity of the image information or the signal-to-noise ratio of the environmental parameters in the original data, and to compare the analysis results generated by the analysis model execution unit with preset agricultural activity logs or historical traceability data to determine the cross-validation consistency weight.

5. The fresh agricultural product end-to-end traceability management system according to claim 1, characterized in that, The data writing scheduling module includes: A threshold judgment unit is used to compare the received confidence score with the high confidence threshold and the low confidence threshold to generate a judgment result; The instruction generation unit is used to generate, based on the judgment result, the write instruction to write the analysis result into the main chain or the side chain to be verified.

6. The fresh agricultural product end-to-end traceability management system according to claim 5, characterized in that, The instruction generation unit is specifically used to generate a status label in the write instruction when the judgment result is to write the analysis result into the side chain to be verified. The status label is used to characterize the pending status of the analysis result, which includes a pending manual review status or a high-risk conflict alarm status.

7. The fresh agricultural product end-to-end traceability management system according to claim 1, characterized in that, The hierarchical blockchain storage module includes: The main chain storage unit is used to store the analysis results that the data writing scheduling module determines should be written to the main chain, so as to form an immutable traceability record; A sidechain storage unit is used to store the analysis results that the data writing scheduling module determines should be written to the sidechain to be verified, so as to isolate the traceability information to be verified. The dynamic verification processing unit is used to update or anchor the status of the analysis results in the main chain storage unit after receiving an external audit confirmation instruction for the analysis results stored in the side chain storage unit.

8. The fresh agricultural product end-to-end traceability management system according to claim 7, characterized in that, The dynamic verification processing unit is specifically used for: Receive the external audit confirmation instruction, which includes authorization verification information and an identifier corresponding to a certain analysis result in the sidechain storage unit; The authorization verification information is validated to confirm the legality of the external audit confirmation instruction; After the authorization verification information is verified, a verification transaction is generated, which includes the identifier and the confirmed status. The verification transaction is submitted to the main chain storage unit so that the traceability status of the analysis result corresponding to the identifier is updated to confirmed.

9. A method for full-chain traceability management of fresh agricultural products, as described in any one of claims 1-8, characterized in that, Includes the following steps: S1. Using the data acquisition module, collect raw data that characterizes the state of fresh agricultural products in the traceability process. S2. Using an artificial intelligence analysis module, receive and process the raw data to generate analysis results and a credibility score corresponding to the analysis results; S3. Using the data writing scheduling module, receive the analysis results and the credibility score, and compare the credibility score with the preset high credibility threshold and low credibility threshold through the smart contract to generate a write instruction for the main chain or the side chain to be verified. S4. Using the hierarchical blockchain storage module, in response to the generated write command, the analysis results are selectively stored in the main chain or the side chain to be verified. S5. When an external audit confirmation instruction is received for the analysis results stored in the sidechain to be verified, the dynamic verification processing unit in the layered blockchain storage module is used to update or anchor the status of the analysis results in the main chain.

10. The method for full-chain traceability management of fresh agricultural products according to claim 9, characterized in that, In step S3, generating write instructions for the main chain or the side chain to be verified specifically includes: When the credibility score is not lower than the high credibility threshold, a write instruction is generated to write the analysis results to the main chain; When the confidence score is lower than the high confidence threshold but not lower than the low confidence threshold, a write instruction is generated to write the analysis result to the sidechain to be verified. When the credibility score is lower than the low credibility threshold, a write instruction is generated to write the analysis results into the sidechain to be verified.

Images