Blockchain-based commodity full-link traceability method and system

Abstract

The present application relates to the technical field of commodity traceability, and in particular to a commodity full-link traceability method and system based on a blockchain. According to state monitoring data and fluctuation degrees of label texts of a commodity to be traced at a target node and a previous link node, and label text association differences between the commodity to be traced and a commodity similar in state to the commodity to be traced at the target node, a semantic anomaly commodity is determined. According to semantic anomaly frequency distribution of the semantic anomaly commodity during a period from a starting point to the target node, and accumulated conditions of physical state fluctuations of the semantic anomaly commodity across nodes during the period, a semantic jump commodity is determined. According to semantic deviation degrees of the semantic jump commodity at adjacent upstream and downstream link nodes of the target node and label category differences of the state similar commodity, a corrected state label of the semantic jump commodity is generated. The present application identifies semantic jumps by longitudinal and lateral state comparison and historical physical fluctuation tracing, reconstructs and corrects labels based on adjacent node data, and reduces traceability evidence distortion.

Description

Technical Field

[0001] This invention relates to the field of commodity traceability technology, specifically to a blockchain-based method and system for full-chain commodity traceability. Background Technology

[0002] In the supply chain of highly sensitive goods such as vaccines, fresh produce, and hazardous chemicals, the physical environment during the flow of goods directly affects product safety and quality. Currently, IoT sensors and blockchain technology are typically combined to monitor and trace highly sensitive goods throughout the entire supply chain. By leveraging the immutability and decentralization of blockchain, the authenticity and reliability of cross-node flow data are ensured, preventing historical records from being tampered with.

[0003] The existing patent document CN121563563A discloses a blockchain-based method and system for tracing the supply chain of canteen ingredients. It identifies and corrects source data anomalies and contamination issues during the supply chain traceability process by analyzing differences in the same attribute parameters of similar ingredients and the correlation between different attribute parameters. The corrected parameter data is then stored on the blockchain for traceability. However, goods undergo multiple physical sorting and information system switches during the actual end-to-end flow. Due to differences in information system coding rules, detection standards, and evaluation dimensions at different nodes, the status labels recorded by each node for the same product often exhibit inconsistent semantic meanings, resulting in semantic jumps in the labels. When the product status label is inconsistent with the upstream, existing methods struggle to effectively distinguish whether the change is a temporary semantic jump caused by node switching or a change in the product's actual physical state. This easily leads to false alarms in traceability or masks genuine anomalies, severely impacting the accuracy and reliability of the blockchain-stored evidence. Summary of the Invention

[0004] To address the technical problem of unreliable traceability and evidence storage caused by the difficulty in distinguishing between temporary semantic jumps in tags and genuine anomalies in products due to differences in node evaluation standards, this invention aims to provide a blockchain-based end-to-end product traceability method. The specific technical solution adopted is as follows: In a first aspect, one embodiment of the present invention provides a blockchain-based method for full-chain traceability of goods, the method comprising: Obtain the original status tags and several status monitoring data of each link node in the circulation path of the product to be traced, and select any intermediate node in the circulation path as the target node; Based on the differences in the degree of fluctuation and the deterioration trend of the various status monitoring data of the traceable product and the same category of products at the target node, the products with similar status at the target node are identified. Based on the fluctuation of the status monitoring data and tag text features of the traceable product at the target node and the previous link node, as well as the difference in tag text association between the traceable product and the product with similar status, the semantically abnormal products at the target node are determined. Based on the frequency distribution of semantic anomalies of semantically abnormal products from the starting point to the target node of the self-circulation path, and the cumulative physical state fluctuations of semantically abnormal products across nodes during this period, the semantically abrupt products at the target node are determined. Based on the semantic deviation of the semantically abruptly changed product from its adjacent upstream and downstream link nodes at the target node and the difference in the types of the original state labels of the similar products, the confidence scores of the upstream and downstream nodes of the semantically abruptly changed product are obtained. Based on the difference between the confidence levels of the upstream and downstream nodes, the original state labels of the adjacent upstream and downstream link nodes are semantically reconstructed to generate corrected state labels for semantically transitioning products.

[0005] Furthermore, determining the status of similar products to the traceable product at the target node includes: Based on the fluctuation and deterioration trend of each status monitoring data of the traceable product at the target node, obtain the status deviation index of each status monitoring data of the traceable product at the target node; During the time the traceable product is at the target node, products that pass through the target node and belong to the same category as the traceable product are identified as reference products for the traceable product at the target node. The absolute values ​​of the differences between the state deviation indices of the traceable product and each of the reference products at the target node are averaged to obtain the comprehensive state difference degree of the two products. From all reference products of the product to be traced at the target node, select reference products whose comprehensive state difference is less than or equal to the preset state difference threshold, and use them as similar products in state to the product to be traced at the target node.

[0006] Furthermore, the step of obtaining the state deviation index of each status monitoring data item of the product to be traced at the target node includes: The difference between each status monitoring data of the traceable product at each link node and the corresponding benchmark monitoring data is taken as the monitoring deviation. The monitoring deviation of the same status monitoring data of the traceable product from the starting point of the circulation path to the target node is arranged in time sequence to obtain the index deviation sequence of each status monitoring data. Obtain the first-order difference sequence of the index deviation sequence, and calculate the ratio of the sum of positive elements in the first-order difference sequence to the sum of the absolute values ​​of all elements, as the deviation growth coefficient for each status monitoring data. Calculate the standard deviation of the indicator deviation sequence, and use the ratio of the monitoring deviation of each status monitoring data of the traceable commodity at the target node to the standard deviation as the relative deviation coefficient of that status monitoring data. The state deviation index of each status monitoring data item of the traceable product at the target node is obtained by multiplying the deviation growth coefficient and the relative deviation coefficient.

[0007] Furthermore, determining the semantically anomalous items of the target node includes: Arrange all status monitoring data of the product to be traced at the target node to obtain the monitoring status vector; perform text vectorization processing on the original status labels of the product to be traced at the target node to obtain the label word vector; Calculate the cosine similarity between the monitoring status vectors of the product to be traced at the target node and the previous link node, and the cosine similarity between the tag word vectors, which are respectively denoted as monitoring status similarity and tag semantic similarity; The homomorphic semantic similarity is obtained by averaging the cosine similarity of the tag word vectors of the traceable product at the target node with all products that are similar to it in all states. Based on the monitoring status similarity, the tag semantic similarity, and the homomorphic semantic similarity, the semantic anomaly of the product to be traced at the target node is obtained; the monitoring status similarity is positively correlated with the semantic anomaly, and the tag semantic similarity and the homomorphic semantic similarity are both negatively correlated with the semantic anomaly. Products whose semantic anomaly degree is greater than a preset semantic deviation threshold are identified as semantically abnormal products of the target node.

[0008] Furthermore, determining the semantically transitioning product of the target node includes: For semantically anomalous products at the target node, the actual state anomalousness of the semantically anomalous products at the target node is obtained based on the degree of state fluctuation of the semantically anomalous products at the adjacent upstream and downstream link nodes of the target node, and the semantic deviation trend of the semantically anomalous products at the downstream link nodes of the target node. The percentage of nodes that were not identified as semantically abnormal products in the link nodes that the semantically abnormal products passed through from the starting point of the flow path to the target node is recorded as the historical jump sparsity; and the arithmetic mean of the actual state abnormality corresponding to the semantically abnormal products being identified as semantically abnormal products in the link nodes they passed through is calculated as the historical state abnormality. A negative correlation mapping is performed on the historical state anomaly degree, and the product of the mapping result and the historical jump sparsity is used as the semantic jump probability degree of the semantically abnormal product of the target node. Products with semantic aberration probability greater than a preset semantic aberration threshold are designated as semantic aberration products for the target node.

[0009] Furthermore, obtaining the substantive state anomaly degree of the semantic jump product of the target node includes: Calculate the absolute value of the difference between the state deviation index of the semantically abnormal product at the target node and each state monitoring data item under its previous link node and next link node, and use the sum of the two absolute values ​​of the difference as the local volatility of each state monitoring data item of the semantically abnormal product at the target node. The cross-node volatility is obtained by averaging the local volatility of all item status monitoring data for semantically abnormal products under the target node. The arithmetic mean of the semantic anomalousness of all state-similar products of the semantically anomalous product during the period from the target node to the next time it is identified as a semantically anomalous product is obtained, and is denoted as the downstream semantic deviation of the semantically anomalous product of the target node. The product of the downstream semantic deviation and the cross-node volatility is used as the substantive state anomaly degree of the semantically abnormal goods of the target node.

[0010] Furthermore, obtaining the upstream node confidence and downstream node confidence of semantically transitioning products includes: For semantically abrupt products at the target node, a preset deviation smoothing factor is added to the semantic anomalies of the semantically abrupt products at the previous and next link nodes of the target node, and the ratio of the accumulated result of the next link node to the accumulated result of the previous link node is used as the relative semantic deviation. The number of original state labels of semantically jump products and similar products in the previous and next link nodes of the target node are counted separately, and a preset fine-grained smoothing factor is added to each. The ratio of the accumulated result of the previous link node to the accumulated result of the next link node is used as the relative label fine-grainedness. The product of the similarity of the monitoring status of the semantically abruptly changed product at the target node and its upstream link node, the relative semantic deviation, and the relative label fine-grainedness is normalized to obtain the upstream node confidence of the semantically abruptly changed product at the target node; the difference between the constant 1 and the upstream node confidence is taken as the downstream node confidence of the semantically abruptly changed product at the target node.

[0011] Furthermore, the generation of the corrected status label for the semantically abrupt product includes: Determine whether the absolute value of the difference between the confidence of the upstream node and the confidence of the downstream node is less than or equal to a preset correction threshold. If so, merge the original state labels of the semantic jump product in the previous link node and the next link node of the target node to generate the corrected state label of the semantic jump product of the target node. If not, the original state label of the link node corresponding to the maximum value of the upstream node confidence score and the downstream node confidence score is used as the corrected state label of the semantic jump product of the target node.

[0012] Furthermore, the Word2Vec algorithm is used to perform text vectorization processing on the original status labels of the traceable products at the target node.

[0013] Secondly, another embodiment of the present invention provides a blockchain-based end-to-end product traceability system, the system comprising: The data acquisition module is used to acquire the original status tags and several status monitoring data of each link node in the circulation path of the traceable product, and to select any intermediate node in the circulation path as the target node. The similar product filtering module is used to determine similar products in the target node based on the differences in the fluctuation and deterioration trend of the various status monitoring data of the product to be traced and the same category of products under the target node. The semantic anomaly identification module is used to determine the semantically abnormal products of the target node based on the fluctuation of the status monitoring data and label text features of the product to be traced at the target node and the previous link node, as well as the difference in label text association between the product to be traced and the product with similar status. The semantic jump identification module is used to determine the semantic jump products at the target node based on the frequency distribution of semantic anomalies from the starting point to the target node of the semantic anomaly product's self-circulation path, as well as the cumulative physical state fluctuations of the semantic anomaly product across nodes during this period. The node confidence assessment module is used to obtain the upstream node confidence and downstream node confidence of the semantically jumping product based on the degree of semantic deviation of the semantically jumping product in the adjacent upstream and downstream link nodes of the target node and the difference in the type of the original state label of the product with similar state. The tag semantic reconstruction module is used to semantically reconstruct the original state tags of the adjacent upstream and downstream link nodes based on the difference between the confidence levels of the upstream nodes and the downstream nodes, and generate corrected state tags for semantically transitioning products.

[0014] The present invention has the following beneficial effects: In this embodiment of the invention, based on the differences in the fluctuation and deterioration trends of the status monitoring data between the traceable product and the same category of products at the target node, similar products with consistent risk of damage are selected. By longitudinally comparing the fluctuation of the status monitoring data and tag text features of the traceable product at the target node and the previous link node, and by horizontally combining the differences in tag text association between the traceable product and similar products, occasional tagging errors caused by system switching or subjective standard differences are effectively eliminated, thereby accurately identifying semantically abnormal products at the target node. The occasional occurrence of abnormal tags is assessed by tracing the historical frequency distribution of semantically abnormal products since the starting point of circulation, combined with the physical status of the product across nodes during this period. The system verifies whether substantial damage has occurred due to accumulated state fluctuations, thereby identifying semantically altered products at the target node. This effectively distinguishes between temporary label changes and actual physical damage to the product. Finally, by comparing the semantic deviation of adjacent upstream and downstream nodes and the differences in the types of original state labels for products with similar states, the system assesses data reliability and obtains the confidence levels of upstream and downstream nodes for semantically altered products. Based on these confidence level differences, the system performs adaptive correction and reconstruction on the original state labels of adjacent nodes, generating corrected state labels for the target node. This objectively restores the true state of the product at the current node, effectively solving the problem of distorted traceability and evidence storage data caused by inconsistent evaluation standards across nodes. Attached Figure Description

[0015] Figure 1 A flowchart illustrating the steps of a blockchain-based end-to-end product traceability method provided in one embodiment of the present invention; Figure 2 A flowchart illustrating a method for obtaining similar goods in a given state, as provided in one embodiment of the present invention; Figure 3 This is a flowchart illustrating a method for obtaining semantically variable products according to an embodiment of the present invention. Figure 4 This is a system architecture diagram of a blockchain-based full-chain traceability system for goods, provided as an embodiment of the present invention. Detailed Implementation

[0016] The following description, in conjunction with the accompanying drawings, details the specific solution of a blockchain-based end-to-end product traceability method and system provided by this invention.

[0017] Example 1: This invention proposes a blockchain-based method for end-to-end product traceability. Please refer to [link / reference]. Figure 1 The diagram illustrates a flowchart of a blockchain-based end-to-end product traceability method according to an embodiment of the present invention. The method includes: Step S1: Obtain the original status tags and several status monitoring data of each link node on the circulation path of the product to be traced, and select any intermediate node in the circulation path as the target node.

[0018] The circulation path refers to the complete transfer route that goods take from their origin to their final consumption location. Link nodes are key handover points and distribution stations along the circulation path, such as refrigerated trucks, distribution centers, or cold storage facilities. This information can be obtained through IoT technologies such as RFID, barcode scanning, and GPS. Goods experience dwell time within link nodes due to warehousing, distribution, or loading / unloading operations. This dwell time is the time interval from when the goods enter the target node to when they leave it, and can be obtained by retrieving inbound and outbound timestamps from the business system. The goods to be traced are specific goods currently undergoing full-link traceability testing and label verification.

[0019] Condition monitoring data refers to continuous time-series values ​​reflecting the physical environment of goods, collected in real time by environmental sensors and intelligent devices deployed within transportation vehicles and warehousing nodes. Specifically, condition monitoring data for fresh produce includes temperature and humidity, for vaccines it includes temperature and vibration, and for hazardous materials it includes temperature, pressure, and tilt. Each condition monitoring data point for a traceable product at each node in the supply chain refers to the arithmetic mean of the condition monitoring data collected during the product's stay at that node.

[0020] Original status labels refer to discrete text status descriptions entered by business personnel or generated by the quality inspection system when goods pass through or are handed over to various links. These descriptions include things like "packaging intact", "outer box slightly deformed", or "compliant". They can be retrieved through the data interface of the intelligent warehouse management system connected to each node.

[0021] Step S2: Based on the differences in the fluctuation and deterioration trend of the status monitoring data of the product to be traced and the same category of products at the target node, determine the products with similar status at the target node.

[0022] In cross-node flow, the true physical state of goods is easily influenced by subjective surface-level labels due to differences in evaluation standards across various business systems. This step avoids relying solely on label comparison. Instead, it selects goods with similar states and consistent risk of damage based on the differences in fluctuation and deterioration trends in the status monitoring data between the goods to be traced and the same category of goods at the target node. This establishes an objective physical state reference group for the goods to be traced, providing a reliable horizontal calibration basis for subsequent identification of semantic shifts in labels.

[0023] Step S3: Based on the fluctuation of the status monitoring data and tag text features of the traceable product at the target node and the previous link node, as well as the difference in tag text association between the traceable product and products with similar status, determine the semantically abnormal products at the target node.

[0024] In the vertical time-series dimension, by comparing the fluctuations of status monitoring data and tag text features, anomalies are identified where the physical state is stable but the tag semantics change abruptly. In the horizontal group dimension, based on the differences in the tag text association between the traceable product and products with similar states, it is determined whether the current abnormal tag belongs to an isolated marker that deviates from the group characteristics. The above multi-dimensional feature cross-comparison effectively isolates occasional tagging errors caused by system switching or subjective standard differences, thereby accurately identifying semantically abnormal products at the target node and narrowing the scope of anomaly investigation for the entire traceability chain.

[0025] Step S4: Based on the frequency distribution of semantic anomalies of semantically abnormal products from the starting point to the target node of the self-flow path, and the cumulative physical state fluctuation of semantically abnormal products across nodes during this period, determine the semantically jump products of the target node.

[0026] By tracing the historical frequency distribution of semantically anomalous products since their origin, the sporadic occurrence of their anomalous labels is assessed. Simultaneously, by considering the cumulative physical state fluctuations of the product across nodes during this period, it is verified whether substantial damage has occurred, thereby identifying the truly semantically anomalous products in the target node. This step effectively distinguishes between temporary label changes and genuine product damage, avoiding system false alarms that could interfere with or contaminate traceability and evidence storage data.

[0027] Step S5: Based on the semantic deviation of the semantically abruptly changed product from its adjacent upstream and downstream link nodes at the target node and the difference in the types of the original state labels of products with similar states, obtain the upstream node confidence and downstream node confidence of the semantically abruptly changed product.

[0028] Given the differences in recording standards among the various business systems, this step assesses the reliability of the output information from each system by comparing the semantic deviation between adjacent upstream and downstream links and the differences in the types of original state tags for products with similar states. This process obtains the confidence levels of semantically abruptly changed products at the upstream and downstream nodes of the target node. This step provides an objective basis for tag data acceptance, effectively avoiding secondary information pollution caused by the introduction of low-quality data during tag correction, and providing reliable support for the subsequent accurate reconstruction of the true state of products.

[0029] Step S6: Based on the difference between the confidence levels of upstream and downstream nodes, semantically reconstruct the original state labels of adjacent upstream and downstream link nodes to generate corrected state labels for semantically transitioned products.

[0030] The difference between the confidence levels of upstream and downstream nodes represents the relative deviation in the reliability of data between two adjacent nodes. Based on the difference, an adaptive reconstruction mechanism is adopted to correct and repair the abnormal original state labels of the target node and generate corrected state labels, so as to objectively restore the true state of the product at the target node. This effectively solves the problem of data distortion in traceability and evidence storage caused by inconsistent evaluation standards across nodes.

[0031] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining goods with similar states is described in [reference needed]. Figure 2 The diagram illustrates a flowchart of a method for obtaining similar goods in a given state, according to an embodiment of the present invention. The method includes: Step S210: Based on the fluctuation and deterioration trend of each status monitoring data of the product to be traced at the target node, obtain the status deviation index of each status monitoring data of the product to be traced at the target node.

[0032] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the state deviation index includes: taking the difference between each state monitoring data of the traceable commodity at each link node and the corresponding benchmark monitoring data as the monitoring deviation amount; arranging the monitoring deviation amounts of the same state monitoring data of the link nodes traversed by the traceable commodity from the starting point of the circulation path to the target node in chronological order to obtain the index deviation sequence of each state monitoring data; obtaining the first-order difference sequence of the index deviation sequence, calculating the ratio of the sum of the positive elements in the first-order difference sequence to the sum of the absolute values ​​of all elements, as the deviation growth coefficient of each state monitoring data; calculating the standard deviation of the index deviation sequence, and taking the ratio of the monitoring deviation amount of each state monitoring data of the traceable commodity at the target node to the standard deviation as the relative deviation coefficient of that state monitoring data; and obtaining the state deviation index of each state monitoring data of the traceable commodity at the target node based on the product of the deviation growth coefficient and the relative deviation coefficient.

[0033] It should be noted that the monitoring deviation reflects the extent to which the actual state of the traceable goods deviates from the normal level at each link node; the indicator deviation sequence records the historical evolution trajectory of the physical state of the traceable goods since shipment, showing its deviation from the normal level. In the first-order difference sequence, positive elements represent an increasing deviation of the state monitoring data from the normal level, i.e., the state is deteriorating. To quantify the proportion of this deteriorating trend in the overall fluctuation, a deviation growth coefficient is obtained. The larger the value, the more significant the cumulative trend of the state monitoring data in the direction of deterioration, and the more continuously the goods are deteriorating. The standard deviation reflects the baseline of the overall state fluctuation of the traceable goods during their historical circulation. The larger the relative deviation coefficient, calculated by the ratio of the monitoring deviation to the standard deviation, the larger it is, indicating that the physical state of the traceable goods at the target node not only deviates from the safety benchmark but also far exceeds its own historical normal fluctuation range, constituting a sudden and severe anomaly. The state deviation index integrates two dimensions: the temporal deterioration trend and the magnitude of the anomaly. The larger its value, the higher the probability that various state monitoring data have experienced real physical anomalies.

[0034] In this embodiment of the invention, historical traceability orders with the same category as the product to be traced and whose final status is marked as normally completed are retrieved from the system database; the average value of each status monitoring data of these historical traceability orders along the circulation path is calculated as the benchmark monitoring data of the product to be traced in that status monitoring data.

[0035] It should be noted that if all elements in the indicator deviation sequence of a certain status monitoring data of the product to be traced are the same at the target node, the corresponding status deviation index will be set to zero.

[0036] Step S220: During the time the traceable product is at the target node, products that pass through the target node and belong to the same category as the traceable product are identified as reference products of the traceable product at the target node.

[0037] It should be noted that, in order to ensure a fair comparison, interference from different environments or the inherent attributes of different products must be eliminated. Therefore, products that pass through the target node during the stay period and belong to the same category as the product to be traced will be identified as reference products.

[0038] Step S230: Average the absolute values ​​of the differences between the state deviation indices of the traceable product and each of its reference products at the target node for all state monitoring data to obtain the comprehensive state difference degree of the two products; select reference products whose comprehensive state difference degree is less than or equal to the preset state difference threshold from all reference products of the traceable product at the target node as similar products in state of the traceable product at the target node.

[0039] It should be noted that the overall health difference degree reflects the degree of deviation between the traceable product and each reference product in terms of overall health status. The smaller the value, the more consistent the abnormal fluctuations and deterioration trends of the traceable product and the reference product in various status monitoring indicators, and the more similar the health status of the two products. Similar status products are those that are in a similar health status to the traceable product at the target node.

[0040] In this embodiment of the invention, a large number of products of the same category at the same node under historical normal circulation records are extracted from the system database to form a historical difference sample set. The mean and standard deviation of the comprehensive state difference in the historical difference sample set are calculated, and the sum of the mean and three times the standard deviation is used as the preset state difference threshold.

[0041] It is important to note that if the product to be traced has no reference product or product in a similar state at the target node, then in the subsequent analysis, both the homomorphic semantic similarity and the relative label fine-grainedness will be set to 1 by default, which means that there is a good-faith assumption that there is no group deviation and that the upstream and downstream characterization granularity is balanced. At the same time, the downstream semantic deviation will be set to 0 by default, which means that there is no downstream group deterioration to confirm it in the absence of corroborating evidence, in order to prevent the chain reaction of misjudgments caused by the lack of reference objects. The units of the default values ​​will be consistent with the units of the respective indicators.

[0042] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining semantically anomalous products includes: arranging all state monitoring data of the product to be traced at the target node to obtain a monitoring state vector; performing text vectorization processing on the original state labels of the product to be traced at the target node to obtain label word vectors; calculating the cosine similarity between the monitoring state vectors of the product to be traced at the target node and the previous link node, and the cosine similarity between the label word vectors, respectively denoted as monitoring state similarity and label semantic similarity; averaging the cosine similarity of the label word vectors of the product to be traced at the target node with all products with similar states to obtain homomorphic semantic similarity; obtaining the semantic anomalousness of the product to be traced at the target node based on the monitoring state similarity, label semantic similarity, and homomorphic semantic similarity; the monitoring state similarity is positively correlated with the semantic anomalousness, while the label semantic similarity and homomorphic semantic similarity are both negatively correlated with the semantic anomalousness; and identifying the product to be traced with a semantic anomalousness greater than a preset semantic deviation threshold as a semantically anomalous product at the target node. In this embodiment, the Word2Vec algorithm is used to perform text vectorization processing on the original status labels of the traceable goods at the target node.

[0043] It should be noted that before calculating the cosine similarity of the monitoring state vectors of the traceable product at the target node and the previous link node, for each status monitoring data of the traceable product at the link node, the Z-score standardization method is used to standardize the monitoring data of that status in the monitoring state vector based on the mean and standard deviation of historical traceability orders of the same category as the traceable product and whose final status was marked as normal completion. This eliminates the dimensional differences between different status monitoring indicators. The monitoring state similarity reflects the stability of the underlying physical state of the product. The higher the value, the more likely that the product has not experienced real physical deterioration during cross-node circulation. In this case, if the recorded label of the product is in stark contrast to the previous link node (i.e., the lower the label semantic similarity), and is severely out of sync with the labels of groups with similar health states (i.e., the lower the homomorphic semantic similarity), it indicates that the change in the current label completely lacks support from physical data and group performance, belonging to a typical isolated false anomaly. The probability of the traceable product encountering a label semantic labeling error at the current node is higher. Therefore, the monitoring state similarity is positively correlated with the semantic anomaly degree, while the label semantic similarity and homomorphic semantic similarity are both negatively correlated with the semantic anomaly degree.

[0044] In this embodiment of the invention, the cosine similarity ranges from -1 to 1. To avoid negative values ​​from participating in subsequent product operations and causing abnormal evaluation logic reversal and distortion, the above three similarities need to be uniformly mapped to the interval of 0 to 1. For the monitoring state similarity, tag semantic similarity, and homomorphic semantic similarity, they are added to constant 1 and then divided by constant 2 to complete the normalization process. The difference between constant 1 and the normalized tag semantic similarity, and the difference between constant 1 and the normalized homomorphic semantic similarity are calculated, and the product of the above two differences and the monitoring state similarity is used as the semantic anomaly degree of the product to be traced at the target node.

[0045] In this embodiment of the invention, a normal sample set is constructed by collecting the semantic anomalies of various products whose label status has been manually verified or confirmed by the system as normal from historical traceability business data. The specific percentile value of the semantic anomaly data distribution in the normal sample set, such as the 95th percentile, is determined as the preset semantic deviation threshold.

[0046] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining semantically variable products can be found in [reference needed]. Figure 3 The diagram illustrates a flowchart of a method for obtaining semantically variable products according to an embodiment of the present invention. The method includes: Step S410: For semantically anomalous products of the target node, obtain the actual state anomalousness of the semantically anomalous products of the target node based on the state fluctuation degree of the semantically anomalous products in the adjacent upstream and downstream link nodes of the target node, and the semantic deviation trend of the semantically anomalous products in the downstream link nodes of the target node.

[0047] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the substantial state anomaly degree includes: calculating the absolute value of the difference between the state deviation index of each state monitoring data item of the semantically anomalous product at the target node and its previous link node and next link node, and taking the sum of the two absolute values ​​of the difference as the local volatility of each state monitoring data item of the semantically anomalous product at the target node; averaging the local volatility of all state monitoring data items of the semantically anomalous product at the target node to obtain the cross-node volatility; obtaining the arithmetic mean of the semantic anomaly degree of all state-similar products of the semantically anomalous product from the target node to the next time it is identified as a semantically anomalous product, and recording it as the downstream semantic deviation degree of the semantically anomalous product of the target node; and taking the product of the downstream semantic deviation degree and the cross-node volatility as the substantial state anomaly degree of the semantically anomalous product of the target node.

[0048] It should be noted that local volatility characterizes the degree of abrupt change in each state monitoring data point at the target node. A higher value indicates more drastic fluctuations in the data point as it passes through the target node. Cross-node volatility is the average of the local volatility of all state monitoring data points, used to characterize the impact on the overall state of semantically anomalous goods. A higher value indicates that semantically anomalous goods encounter multi-dimensional environmental shocks at the target node, increasing the risk of substantial damage. Downstream semantic deviation provides evidence of the subsequent evolution of the goods' state. A higher value indicates that similar goods experiencing similar environments continue to exhibit anomalies in downstream circulation, and the greater the likelihood that the harsh environment at the target node has indeed caused substantial damage, as evidenced by the group evolution trend. Therefore, both downstream semantic deviation and cross-node volatility are positively correlated with the degree of substantial state anomaly. A higher degree of substantial state anomaly reflects a higher risk of irreversible, real-state damage to the goods, rather than temporary labeling errors due to node handover or recording differences.

[0049] It should be noted that if no node is identified as a semantically abnormal product again from the target node to the end of the flow path, the end point is used as the termination boundary. The arithmetic mean of the semantic abnormality of all similar products in the link nodes passed by the semantically abnormal product from the target node to the end of the flow is obtained as the downstream semantic deviation.

[0050] Step S420: Obtain the percentage of nodes that are not identified as semantically anomalous products in the link nodes traversed from the starting point of the semantically anomalous product's self-flow path to the target node, denoted as the historical jump sparsity; calculate the arithmetic mean of the actual state anomalies corresponding to when the semantically anomalous product is identified as a semantically anomalous product in the traversed link nodes, as the historical state anomaly; perform a negative correlation mapping on the historical state anomalies, and use the product of the mapping result and the historical jump sparsity as the semantic jump probability of the semantically anomalous product of the target node; identify semantically anomalous products with a semantic jump probability greater than a preset semantic jump threshold as semantically jump products of the target node.

[0051] It should be noted that if a product has very few abnormal records during its historical circulation, and past occasional abnormalities have not been accompanied by substantial deterioration of its state, then the labeling abnormality appearing at the current node is highly likely a temporary jump caused by inconsistent detection standards across nodes. Historical state anomaly degree characterizes the actual degree of damage to a semantically abnormal product when it is identified as abnormal during historical circulation. The smaller the value, the more likely that the product did not experience substantial state damage in past historical warnings, and that past abnormalities were mostly temporary labeling errors. The greater the historical jump sparsity, the better the semantically abnormal product's state assessment was in most historical nodes before the target node, with fewer abnormal records. Therefore, historical state anomaly degree is negatively correlated with semantic jump probability, while historical jump sparsity is positively correlated with semantic jump probability. The greater the semantic jump probability, the greater the likelihood that the product's state abnormality at the target node is a temporary semantic jump caused by information gaps or recording errors during circulation, rather than irreversible substantial state damage.

[0052] In this embodiment of the invention, the negative number of the historical state anomaly degree of the semantically abnormal goods of the target node is used as the exponent of an exponential function with the natural constant as the base to achieve negative correlation mapping.

[0053] In this embodiment of the invention, a first type of product sample that has been identified as a temporary system labeling error and has not suffered actual physical damage is extracted from the historical circulation records, and the mean of the semantic jump probability corresponding to the first type of product sample is calculated; at the same time, a second type of product sample that has been identified as having actual physical damage is extracted from the historical records, and the mean of the semantic jump probability corresponding to the second type of product sample is calculated; and the arithmetic mean of the above two means is determined as the preset semantic jump threshold.

[0054] It should be noted that if the number of links through which a semantically abnormal product is identified as a semantically abnormal product in the self-flow path from the starting point to the target node is zero, then the historical state abnormality degree is set to 0, and the dimension is the same as the historical state abnormality degree.

[0055] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the confidence of the upstream node and the confidence of the downstream node includes: for the semantically abrupt change of the product at the target node, adding a preset deviation smoothing factor to the semantic anomaly of the product at the previous link node and the next link node of the target node respectively, and taking the ratio of the accumulated result of the next link node to the accumulated result of the previous link node as the relative semantic deviation; counting the number of original state labels of the product with similar state at the previous link node and the next link node of the target node respectively, and adding a preset fine-grained smoothing factor to each, and taking the ratio of the accumulated result of the previous link node to the accumulated result of the next link node as the relative label fine-grainedness; normalizing the product of the monitored state similarity, relative semantic deviation, and relative label fine-grainedness of the product at the target node and its previous link node to obtain the upstream node confidence of the product with the semantically abrupt change of the target node; and taking the difference between the constant 1 and the upstream node confidence as the downstream node confidence of the product with the semantically abrupt change of the target node.

[0056] In one specific implementation of this invention, the confidence level of the upstream node is expressed by the formula: In the formula, The confidence level of the upstream node for the semantically transitioning product of the target node; The semantic anomaly degree of the semantic jump product of the target node in the previous link node of the target node; The semantic anomaly degree of the semantically transitioned product at the target node in the next link node; The number of types of original state labels of products that are similar to the target node's semantic jump product in the previous link node of the target node; The number of original state labels of similar products in the next link node of the target node for the semantically changed product of the target node; The similarity of the monitoring status of semantically transitioning products at the target node to that of the upstream link node. To preset the deviation smoothing factor, To predefine fine-grained smoothing factors, the two smoothing factors are used to prevent fractions from becoming invalid due to a zero denominator. Since the semantic anomaly degree ranges from 0 to 1, then... The value is 0.001. Since the number of original status label categories is an integer, then... The value is 1; The relative semantic deviation of the semantically transitioning product at the target node; The relative label of the semantically transitioning product at the target node is a fine-grained value; Norm is a normalization function, preferably a proportional mapping function. Where X is the product of relative semantic deviation, relative label fineness, and monitoring state similarity, this function can smoothly map an unbounded positive real number X to... Within the interval, and when the overall performance of upstream and downstream nodes is evenly matched (i.e., X=1), an adaptive output of a balanced confidence level of 0.5 can be achieved.

[0057] It should be noted that when a target node's label becomes invalid due to a semantic shift, adjacent nodes need to be introduced to perform corrective reconstruction. If the physical state of the preceding node is more consistent, the records are more rigorous, and the characterization is more detailed, then it should be given a higher confidence weight. This is because: consistent state confirms that the product entity has not changed, rigorous records verify data reliability, and high granularity ensures sufficient support for reconstruction, thereby effectively guaranteeing that the final generated corrected label accurately matches the product's true state. A greater relative semantic deviation indicates that the semantic anomaly probability of the preceding node is much lower than that of the following node, and the label records output by its business system are more rigorous and reliable; a greater relative label granularity indicates that the preceding node's classification and characterization of the product state is more refined and comprehensive, and the data reference value is higher; monitoring state similarity... The higher the confidence level, the more consistent the actual physical state of the traceable product at the target node is with the previous node in the chain, indicating a very strong applicability of past labels. Therefore, the confidence level of the upstream node... The larger the value, the better the overall information quality of the previous link node, and the more data dominance it should have when generating corrected status labels.

[0058] It is important to note that when the target node is the starting point of the flow path, the confidence level of the upstream node is directly assigned to 0, and the confidence level of the downstream node is assigned to 1. When the target node is the ending point of the flow path, the confidence level of the upstream node is directly assigned to 1, and the confidence level of the downstream node is assigned to 0. When the target node is an intermediate node but its upstream link node is the starting point of the flow path, the semantic anomaly level of the upstream link node is set to 0 by default. When the next link node of the target node is the ending point of the flow path, the semantic anomaly level of the next link node is set to 0 by default.

[0059] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the corrected state label includes: determining whether the absolute value of the difference between the confidence of the upstream node and the confidence of the downstream node is less than or equal to a preset correction threshold; if so, then fusing the original state labels of the semantically changed product in the previous link node and the next link node of the target node to generate the corrected state label of the semantically changed product of the target node; if not, then using the original state label of the link node corresponding to the maximum value of the confidence of the upstream node and the confidence of the downstream node as the corrected state label of the semantically changed product of the target node.

[0060] It should be noted that when the absolute value of the difference between the confidence levels of upstream and downstream nodes is less than or equal to a preset correction threshold, it indicates that the information reliability of the upstream and downstream links of the target node is relatively balanced. Discarding the label of either node alone may result in the omission of crucial status information. Therefore, by using string concatenation and other methods, the original status labels of the upstream and downstream links are merged to generate corrected status labels, ensuring the integrity of traceability information. Conversely, it indicates that the information quality of one end of the upstream and downstream links of the target node is significantly superior, while the other end may have data contamination or recording bias. To prevent low-quality data from causing secondary interference to the final traceability results, the original status label of the node with the highest confidence is extracted as the corrected status label, thereby effectively isolating erroneous information. Through the above adaptive diversion mechanism, semantic jump errors occurring at the target node are corrected, and the finally generated corrected status labels are packaged and uploaded to the blockchain for evidence storage, ensuring the objectivity, authenticity, and immutability of the entire product traceability data.

[0061] In this embodiment of the invention, the arithmetic mean of the normalized results of the similarity of the monitoring status of the semantic jump product of the target node with its previous link node and the next link node is calculated, and the product of the arithmetic mean and a set baseline constant (such as 0.2) is determined as the preset correction threshold.

[0062] This invention is now complete.

[0063] Example 2: This invention proposes a blockchain-based end-to-end product traceability system. Please refer to [link / reference]. Figure 4 The diagram illustrates a system architecture of a blockchain-based end-to-end product traceability system according to an embodiment of the present invention. The system includes: The data acquisition module is used to acquire the original status tags and several status monitoring data of each link node in the circulation path of the traceable product, and to select any intermediate node in the circulation path as the target node. The similar product filtering module is used to determine similar products in the target node based on the differences in the fluctuation and deterioration trend of the various status monitoring data of the product to be traced and the same category of products under the target node. The semantic anomaly identification module is used to identify semantically abnormal products at the target node based on the fluctuation of the status monitoring data and label text features of the product to be traced at the target node and the previous link node, as well as the difference in the label text association between the product to be traced and the product with similar status. The semantic jump identification module is used to determine the semantic jump products at the target node based on the frequency distribution of semantic anomalies from the starting point to the target node of the semantic anomaly product's self-circulation path, as well as the cumulative physical state fluctuation of the semantic anomaly product across nodes during this period. The node confidence assessment module is used to obtain the upstream node confidence and downstream node confidence of semantically jump products based on the degree of semantic deviation of semantically jump products in the adjacent upstream and downstream link nodes of the target node and the difference in the types of original state labels of products with similar states. The tag semantic reconstruction module is used to semantically reconstruct the original state tags of adjacent upstream and downstream link nodes based on the difference between the confidence levels of upstream and downstream nodes, and generate corrected state tags for semantically transitioning products.

[0064] It should be noted that the devices provided in the above embodiments are only illustrative examples of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above. Furthermore, the blockchain-based full-chain traceability system for goods and the blockchain-based full-chain traceability method embodiment provided in the above embodiments belong to the same concept, and their specific implementation process is detailed in the method embodiment, which will not be repeated here.

Claims

1. A blockchain-based method for full-chain traceability of goods, characterized in that, The method includes: Obtain the original status tags and several status monitoring data of each link node in the circulation path of the product to be traced, and select any intermediate node in the circulation path as the target node; Based on the differences in the degree of fluctuation and the deterioration trend of the various status monitoring data of the traceable product and the same category of products at the target node, the products with similar status at the target node are identified. Based on the fluctuation of the status monitoring data and tag text features of the traceable product at the target node and the previous link node, as well as the difference in tag text association between the traceable product and the product with similar status, the semantically abnormal products at the target node are determined. Based on the frequency distribution of semantic anomalies of semantically abnormal products from the starting point to the target node of the self-circulation path, and the cumulative physical state fluctuations of semantically abnormal products across nodes during this period, the semantically abrupt products at the target node are determined. Based on the semantic deviation of the semantically abruptly changed product from its adjacent upstream and downstream link nodes at the target node and the difference in the types of the original state labels of the similar products, the confidence scores of the upstream and downstream nodes of the semantically abruptly changed product are obtained. Based on the difference between the confidence levels of the upstream and downstream nodes, the original state labels of the adjacent upstream and downstream link nodes are semantically reconstructed to generate corrected state labels for semantically transitioning products.

2. The blockchain-based end-to-end product traceability method according to claim 1, characterized in that, The process of determining similar products in the target node to the product to be traced includes: Based on the fluctuation and deterioration trend of each status monitoring data of the traceable product at the target node, obtain the status deviation index of each status monitoring data of the traceable product at the target node; During the time the traceable product is at the target node, products that pass through the target node and belong to the same category as the traceable product are identified as reference products for the traceable product at the target node. The absolute values ​​of the differences between the state deviation indices of the traceable product and each of the reference products at the target node are averaged to obtain the comprehensive state difference degree of the two products. From all reference products of the product to be traced at the target node, select reference products whose comprehensive state difference is less than or equal to the preset state difference threshold, and use them as similar products in state to the product to be traced at the target node.

3. The blockchain-based end-to-end product traceability method according to claim 2, characterized in that, The process of obtaining the state deviation index of each status monitoring data item for the product to be traced at the target node includes: The difference between each status monitoring data of the traceable product at each link node and the corresponding benchmark monitoring data is taken as the monitoring deviation. The monitoring deviation of the same status monitoring data of the traceable product from the starting point of the circulation path to the target node is arranged in time sequence to obtain the index deviation sequence of each status monitoring data. Obtain the first-order difference sequence of the index deviation sequence, and calculate the ratio of the sum of positive elements in the first-order difference sequence to the sum of the absolute values ​​of all elements, as the deviation growth coefficient for each status monitoring data. Calculate the standard deviation of the indicator deviation sequence, and use the ratio of the monitoring deviation of each status monitoring data of the traceable commodity at the target node to the standard deviation as the relative deviation coefficient of that status monitoring data. The state deviation index of each status monitoring data item of the traceable product at the target node is obtained by multiplying the deviation growth coefficient and the relative deviation coefficient.

4. The blockchain-based end-to-end product traceability method according to claim 3, characterized in that, The determination of semantically abnormal products for the target node includes: Arrange all status monitoring data of the product to be traced at the target node to obtain the monitoring status vector; perform text vectorization on the original status labels of the product to be traced at the target node to obtain the label word vector. Calculate the cosine similarity between the monitoring status vectors of the product to be traced at the target node and the previous link node, and the cosine similarity between the tag word vectors, which are respectively denoted as monitoring status similarity and tag semantic similarity; The homomorphic semantic similarity is obtained by averaging the cosine similarity of the tag word vectors of the traceable product at the target node with all products that are similar to it in all states. Based on the monitoring status similarity, the tag semantic similarity, and the homomorphic semantic similarity, the semantic anomaly of the product to be traced at the target node is obtained; the monitoring status similarity is positively correlated with the semantic anomaly, and the tag semantic similarity and the homomorphic semantic similarity are both negatively correlated with the semantic anomaly. Products whose semantic anomaly degree is greater than a preset semantic deviation threshold are identified as semantically abnormal products of the target node.

5. The blockchain-based end-to-end product traceability method according to claim 4, characterized in that, The process of determining the semantically transitioning product of the target node includes: For semantically anomalous products at the target node, the actual state anomalousness of the semantically anomalous products at the target node is obtained based on the degree of state fluctuation of the semantically anomalous products at the adjacent upstream and downstream link nodes of the target node, and the semantic deviation trend of the semantically anomalous products at the downstream link nodes of the target node. The percentage of nodes that were not identified as semantically abnormal products in the link nodes that the semantically abnormal products passed through from the starting point of the self-flow path to the target node is recorded as the historical jump sparsity; and the arithmetic mean of the actual state abnormality corresponding to the semantically abnormal products being identified as semantically abnormal products in the link nodes they passed through is calculated as the historical state abnormality. A negative correlation mapping is performed on the historical state anomaly degree, and the product of the mapping result and the historical jump sparsity is used as the semantic jump probability degree of the semantically abnormal product of the target node. Products with semantic aberration probability greater than a preset semantic aberration threshold are designated as semantic aberration products for the target node.

6. The blockchain-based end-to-end product traceability method according to claim 5, characterized in that, The process of obtaining the substantive state anomaly degree of the semantic jump product of the target node includes: Calculate the absolute value of the difference between the state deviation index of the semantically abnormal product at the target node and each state monitoring data item under its previous link node and next link node, and use the sum of the two absolute values ​​of the difference as the local volatility of each state monitoring data item of the semantically abnormal product at the target node. The cross-node volatility is obtained by averaging the local volatility of all item status monitoring data for semantically abnormal products under the target node. The arithmetic mean of the semantic anomalousness of all state-similar products of the semantically anomalous product during the period from the target node to the next time it is identified as a semantically anomalous product is obtained, and is denoted as the downstream semantic deviation of the semantically anomalous product of the target node. The product of the downstream semantic deviation and the cross-node volatility is used as the substantive state anomaly degree of the semantically abnormal goods of the target node.

7. The blockchain-based end-to-end product traceability method according to claim 4, characterized in that, The process of obtaining the upstream node confidence score and downstream node confidence score of semantically abruptly changing products includes: For semantically abrupt products at the target node, a preset deviation smoothing factor is added to the semantic anomalies of the semantically abrupt products at the previous and next link nodes of the target node, and the ratio of the accumulated result of the next link node to the accumulated result of the previous link node is used as the relative semantic deviation. The number of original state labels of semantically jump products and similar products in the previous and next link nodes of the target node are counted separately, and a preset fine-grained smoothing factor is added to each. The ratio of the accumulated result of the previous link node to the accumulated result of the next link node is used as the relative label fine-grainedness. The product of the similarity of the monitoring status of the semantically abruptly changed product at the target node and its upstream link node, the relative semantic deviation, and the relative label fine-grainedness is normalized to obtain the upstream node confidence of the semantically abruptly changed product at the target node; the difference between the constant 1 and the upstream node confidence is taken as the downstream node confidence of the semantically abruptly changed product at the target node.

8. The blockchain-based end-to-end product traceability method according to claim 1, characterized in that, The generation of the corrected status label for the semantically abrupt product includes: Determine whether the absolute value of the difference between the confidence of the upstream node and the confidence of the downstream node is less than or equal to a preset correction threshold. If so, merge the original state labels of the semantic jump product in the previous link node and the next link node of the target node to generate the corrected state label of the semantic jump product of the target node. If not, the original state label of the link node corresponding to the maximum value of the upstream node confidence score and the downstream node confidence score is used as the corrected state label of the semantic jump product of the target node.

9. The blockchain-based end-to-end product traceability method according to claim 4, characterized in that, The Word2Vec algorithm is used to perform text vectorization of the original status labels of the products to be traced at the target node.

10. A blockchain-based end-to-end product traceability system, characterized in that: The system includes: The data acquisition module is used to acquire the original status tags and several status monitoring data of each link node in the circulation path of the traceable product, and to select any intermediate node in the circulation path as the target node. The similar product filtering module is used to determine similar products in the target node based on the differences in the fluctuation and deterioration trend of the various status monitoring data of the product to be traced and the same category of products under the target node. The semantic anomaly identification module is used to determine the semantically abnormal products of the target node based on the fluctuation of the status monitoring data and label text features of the product to be traced at the target node and the previous link node, as well as the difference in label text association between the product to be traced and the product with similar status. The semantic jump identification module is used to determine the semantic jump products at the target node based on the frequency distribution of semantic anomalies from the starting point to the target node of the semantic anomaly product's self-circulation path, as well as the cumulative physical state fluctuations of the semantic anomaly product across nodes during this period. The node confidence assessment module is used to obtain the upstream node confidence and downstream node confidence of the semantically abruptly ... The tag semantic reconstruction module is used to semantically reconstruct the original state tags of the adjacent upstream and downstream link nodes based on the difference between the confidence levels of the upstream nodes and the downstream nodes, and generate corrected state tags for semantically transitioning products.

Images