Commodity traceability management method and system based on big data

Abstract

The invention relates to the technical field of big data processing and analysis, in particular to a commodity traceability management method and system based on big data. The method comprises the steps of determining a data range and association logic of each type of events by constructing a semantic event system of a commodity life cycle, performing data acquisition, preprocessing and credibility evaluation at edge nodes and generating an evidence header, and further converting data into multilayer verifiable commodity fingerprints of an L1 edge abstract, L2 event level fusion and L3 full link fusion, the method comprises the following steps: realizing feature fusion and dynamic incremental updating, then performing distributed notarization and multi-party verification on fingerprints through a federated anchor point mechanism to ensure that a full link is traceable and cannot be tampered, and finally supporting full link traceability query and anomaly tracking based on a unique commodity identifier. And closed-loop traceable management is realized through visual display and multi-role authority management. According to the invention, the reliability, interpretability and cross-organization trust of commodity traceability are effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of big data processing and analysis, and particularly relates to a commodity traceability management method and system based on big data. BACKGROUND

[0002] With the acceleration of the globalization and digitization of commodity supply chains, the amount of data generated at each link of the industry chain has increased dramatically and is diverse in modalities, covering multiple stages such as production, logistics, warehousing, and sales. Modern supply chain management has higher requirements for the real-time, completeness, and credibility of traceability information, and the development of Internet of Things, edge computing, and big data technology provides a technical basis for building a fine-grained and verifiable commodity traceability system.

[0003] A traceable management system and method based on a breeding industry service platform are disclosed in Chinese patent application CN120579984A, which includes a server, a mobile data anchoring device, a detachable traceability label module, a life cycle state management module, an event hash and anchor verification module, and a blockchain intelligent verification module. The detachable traceability label module is used to identify the identity and life cycle state of the breeding individual and trigger state change events during the disassembly process. The mobile data anchoring device collects data from key nodes and forms anchor records. The life cycle state management module maintains the life cycle logical state diagram of each individual and checks whether all event uploads are compliant. The event hash and anchor verification module encapsulates all behavior nodes into structured hash events, compares them with the current state diagram for verification, and forms trusted anchors. The blockchain intelligent verification module performs contract-level verification on event hashes and state paths.

[0004] At the same time, existing traceability technologies are gradually developing from centralized records to distributed collaborative verification. How to achieve fine-grained event semantic modeling, edge data credible processing, cross-organizational secure notarization, and full-link dynamic tracking has become a key research direction for improving supply chain traceability capabilities. SUMMARY

[0005] The present application aims to address the problems in the background art and proposes a commodity traceability management method and system based on big data.

[0006] The technical solution of the present application is a commodity traceability management method based on big data, which includes the following specific implementation steps: S1. Construct a semantic event system for the life cycle of commodities, and clearly define the data range, trigger conditions, and association logic with previous and subsequent events for each type of event. S2. Collect, preprocess, and preliminarily assess the credibility of semantic events at the edge node, and generate semantic segmented data and evidence headers. S3. Transform edge-collected data into a multi-layered verifiable product fingerprint that combines L1 edge summarization, L2 event-level fusion, and L3 end-link fusion. S4. Distributed notarization and multi-party verification of the multi-level verifiable product fingerprints are performed through a federated anchor mechanism. S5. Conduct full-chain traceability and anomaly tracking of goods based on multi-level verifiable product fingerprints and federated anchor notarization results.

[0007] Preferably, constructing a semantic event system for the product lifecycle specifically includes: Based on product category, production process characteristics and regulatory requirements, a lifecycle semantic event framework is established, dividing the product lifecycle into several semantic segments with behavioral meaning and process constraints; Define the data types that can be collected, the data source devices, and the responsible parties for collecting data for each semantic event; Establish a micro-condition triggering mechanism for each type of event, and automatically trigger enhanced acquisition strategies based on sensor mutations, image anomalies, positional drift, or missing records. Establish the relationships between events and the fingerprint input strategy, and clarify that the data of the preceding events will be used as contextual information to participate in the fingerprint calculation of subsequent events.

[0008] Preferably, step S2 specifically includes: Automatically generate collection configurations based on event semantics and device capabilities, dynamically adjust sampling modes and frequencies, and realize adaptive access and contextualized sampling strategies for heterogeneous devices. Synchronize time and verify data integrity at edge nodes to generate an evidence header containing device information, sampling period, and local credibility. Perform local credibility scoring and lightweight anomaly detection on the collected data, filter out noise or anomalous data, and trigger enhancement strategies. Data is divided into semantic segments based on event semantics, priority-aware compression is performed, and a layered caching strategy is adopted to ensure the safe and complete reporting of critical data when the network is limited.

[0009] Preferably, step S3 specifically includes: L1 fingerprint is generated by generating an edge summary fingerprint based on semantic segmentation and local evidence header. The L1 fingerprint includes multimodal feature summary, device information, time window and confidence score. Generate L2 fingerprints, fuse all L1 fingerprints and context information within a single event, assign weights based on trustworthiness and enhancement flags, and preserve inter-event dependencies; Generate L3 fingerprints by integrating all event information of the same product to generate a full-chain product fingerprint, and record event sequences, anomaly markers and key features; The L1 fingerprint, L2 fingerprint and L3 fingerprint are subjected to integrity and time continuity check, and are dynamically corrected and incrementally updated in combination with edge caching and enhancement strategy.

[0010] Preferably, the generation of the L1 fingerprint comprises: extracting a key frame feature vector for image data; acquiring a turning point sequence or a feature curve vector for time series data; acquiring a field summary and hash combination for text or structured data; The L1 fingerprint records data integrity and time continuity indicators when generated.

[0011] Preferably, the generation of the L2 fingerprint comprises: A multi-channel weighted fusion method is adopted, and each channel weight is dynamically allocated according to the edge credibility score and the enhancement strategy label; The temporal context is integrated to ensure the time series continuity of the data in the event; The influence of the preceding event is recorded according to the event correlation strategy.

[0012] Preferably, the generation of the L3 fingerprint comprises: An inter-event weight allocation strategy is adopted in combination with micro-condition triggering, abnormality labeling and credibility scoring; A multi-level weighted average and key feature reservation method is adopted for cross-event feature vector fusion; Abnormal information is embedded in the full link fingerprint, recording event abnormalities, enhancement strategy triggering records and paragraphs with low local credibility.

[0013] Preferably, the step S4 specifically comprises: An anchor record is generated after the occurrence of a key event of a commodity, the anchor record contains an event ID, an L2 fingerprint summary or an L3 fingerprint summary, a timestamp, a context label and an edge node signature, and a distributed ledger is formed through multi-node synchronization; Each node receives the anchor point and performs local verification, including fingerprint integrity, time continuity and evidence header consistency check, and generates a node weight according to the verification result; The validity of the anchor point is confirmed according to the node weight and the verification result, the anchor point exceeding the set threshold is recorded to the global link ledger, and the preceding event dependency relationship is updated; The supervisory authority, supply chain participants or third-party audit nodes independently verify the anchor point to check the fingerprint consistency and link integrity.

[0014] Preferably, the step S5 comprises: The user queries the full link event chain through the unique identification of the commodity, and the system integrates the L1 fingerprint, the L2 fingerprint, the L3 fingerprint, the notarization status of the anchor point, the node weight and the event context information to generate a visual link diagram. The system performs multi-dimensional anomaly determination on the full-link event based on edge credibility, event sequence integrity and node weight, automatically locates the abnormal event node and analyzes the abnormal type and influence range; When a new event or an updated event is added in the commodity life cycle, the system only incrementally updates the related L1 fingerprint, L2 fingerprint, L3 fingerprint and anchor chain; The system provides differentiated query permissions according to user roles, and combines anchor notarization to ensure data credibility.

[0015] The technical scheme of the present application: a commodity traceability management system based on big data, which is used for the above-mentioned commodity traceability management method based on big data, comprising: A multi-source data acquisition and preprocessing module is used to collect original commodity data from each link of the supply chain and perform preliminary cleaning, formatting and time sequence correction to generate standardized event records; An edge evidence header generation and credibility enhancement module is used to generate event evidence headers at the acquisition node end, record event time, acquisition device identity and preliminary credibility score, and perform data compression, summarization and credibility enhancement; A multi-layer fingerprint generation and incremental update module is used to extract features from the preprocessed multi-source data and edge evidence headers, generate L1 fingerprint, L2 fingerprint and L3 fingerprint, and support incremental update mechanism; A federal anchor notarization and multi-party verification module is used to perform distributed anchor notarization on the multi-layer fingerprint, synchronize and confirm between federal nodes, and support multi-party verification mechanism; A commodity traceability query and abnormality tracking module is used to provide full-link commodity information query, visual display and multi-dimensional abnormality tracking functions.

[0016] Compared with the prior art, the above technical scheme of the present application has the following beneficial technical effects: This invention designs a product traceability management method and system based on big data. By constructing a semantic event system, it achieves structured collection and association of product lifecycle data, giving the traceability chain clear behavioral meaning and auditability. The edge computing-based data collection and evidence header generation mechanism completes preprocessing and credibility assessment at the data source, effectively improving data quality and on-site processing capabilities. By establishing a multi-level product fingerprint structure from L1 to L3, it achieves feature fusion and dynamic incremental updates of multimodal data, ensuring both the uniqueness and verifiability of traceability identifiers and improving system processing efficiency. The introduced federated anchor notarization mechanism, through distributed node verification and dynamic weight consensus, ensures the immutability of end-to-end data and cross-organizational trust, significantly enhancing the credibility and notarization of traceability results. Finally, by integrating visual query and multi-dimensional anomaly tracking functions, it provides transparent and reliable end-to-end traceability services for regulatory agencies, supply chain participants, and consumers, achieving closed-loop traceability management. Overall, it effectively solves multiple problems in traditional traceability systems, such as data silos, insufficient credibility, low traceability efficiency, and difficulties in cross-organizational collaboration. Attached Figure Description

[0017] Figure 1 This is a flowchart of a product traceability management method based on big data proposed in this invention; Figure 2 This is a system architecture diagram of a big data-based product traceability management system proposed in this invention. Detailed Implementation

[0018] Example 1, as Figure 1 As shown, the present invention proposes a product traceability management method based on big data, which includes the following specific implementation steps: S1. By constructing a semantic event system for the product lifecycle, the scope of data to be collected, triggering conditions, and the logical relationship with preceding and following events for each type of event are clearly defined. This gives the traceability chain structural behavioral meaning and auditability. An executable event model is constructed using four layers of mechanisms: semantic segmentation, responsibility boundaries, micro-condition triggering, and data association. Specifically: S11. Based on the product category, production process characteristics and regulatory requirements, establish a life cycle semantic event framework from a macro perspective. This framework divides the life cycle into several semantic segments with behavioral significance and process constraints through observation of actual operation scenarios in the supply chain, such as "raw material environmental inspection, semi-finished product packaging confirmation, unified review before packing, temperature control calibration at transportation nodes, and warehouse static monitoring". S12. For each semantic event, define the types of data that can be collected, the data source devices and the responsible parties, including but not limited to images, sensor curves, work records and location information, and specify which data must be confirmed on-site and which can be supplemented asynchronously. By setting the "data responsibility boundary", it is ensured that each segment of the traceability data has a traceable source subject, that is, for each semantic event established in step S11, the type of collectable data, the source device and the collection responsibility party are further refined; For example, in the "pre-packing review" event, camera images, packaging label codes, work personnel confirmation information, temperature and humidity readings, etc. are collected as the data range; while in the "transportation node temperature control calibration" event, vehicle GPS, temperature control device operating status, and environmental temperature and humidity curve are the main data sources; S13, a micro-condition triggering mechanism is established for each type of event, and enhanced collection strategies are automatically triggered according to sensor mutations, image abnormalities, position drifts or record missing, such as increasing sampling frequency or recording additional snapshots, so as to capture potential abnormalities without increasing excessive storage, that is, for the data range defined in step S12, a set of "micro-conditions" is set for each event to identify risk points, sudden situations or suspicious features; For example, sudden jumps in temperature sensors, camera detection of packaging integrity abnormalities, short-term GPS drift, repeated or missing work records, etc. are considered as micro-condition triggering points; Once the micro-condition is triggered, the enhanced strategy preset for the event is immediately enabled, such as increasing the number of image captures, increasing the sampling frequency, generating a special marked fingerprint subset, or triggering manual review information supplement; S14, the association relationship between events and the fingerprint input strategy are established, and it is clear which data of the previous event will participate in the fingerprint calculation of the subsequent event, such as using inspection images, temperature control curves or work records as contextual association information, so that the traceability chain has continuity and logical consistency, that is, the association relationship between each event is clear, and the results of which events will be used as contextual information for generating fingerprints of subsequent events; For example, the image features of "raw material warehouse inspection" can be used as historical reference for "semi-finished product packaging confirmation", and the temperature and humidity curve in "transportation temperature control calibration" can be used as associated input for "warehouse standing monitoring".

[0019] S2, the semantic events defined in step S1 are efficiently collected, pre-processed and preliminarily evaluated at the edge node, and semantic segmented data and evidence headers are generated for the fingerprint generator, while the data integrity and auditability are ensured through intelligent compression and offline caching, combining data collection, adaptation, time synchronization, and credibility evaluation into an edge evidence engine, so that the data can be partially intelligently processed on site and the link continuity is ensured, specifically: S21, automatically generate collection configuration according to event semantics and device capabilities, dynamically adjust sampling mode and frequency, and realize adaptive access and situational sampling strategy of heterogeneous devices, specifically: At the specified site of each semantic event, various data sources are accessed to the edge node through a unified access protocol (or adapter), including but not limited to: industrial cameras, temperature / humidity / pressure / vibration sensors, RFID / NFC readers, GPS locators, PLC / SCADA interfaces, and manual terminals (code scanning guns, mobile input); When the device is accessed, the edge node performs capability description exchange to obtain the sampling capability, timestamp precision, minimum available bandwidth, and local storage capability meta information of the device; Based on the device capability and the event semantics defined in step S1, the edge node generates a collection configuration, including but not limited to sampling frequency, sampling mode (continuous, periodic, triggered), compression / format strategy, and priority (which data must be uploaded first, which can be delayed); For example, for industrial cameras, set to "key frame + differential frame" mode for visual review scenarios; for temperature sensors, use "window average + mutation capture" mode; During the collection process, context-adaptive sampling is used, and the edge node dynamically adjusts the sampling rate according to the environment and historical behavior; For example, if the temperature of the packing unit is stable for a short period of time, the sampling rate is reduced, and if slight jitter or abnormality is detected, the sampling frequency is increased for a short period of time and enhanced snapshots are added; S22, synchronize time at the edge node and verify data integrity, generate an evidence header containing device information, sampling period, and local credibility, provide reliable meta information for subsequent fingerprint generation, specifically: Each edge node maintains a double-layer time synchronization strategy, preferentially synchronizes with the upstream gateway through the Network Time Protocol (NTP / PTP), and uses local sliding window correction when the network is unreachable (uses cross-alignment between multiple devices, such as aligning camera frame time and sensor sampling sequence), ensuring time consistency of each channel within the event in the short term; Each time a data segment is collected or packaged, a local evidence header is generated at the edge node, containing the source device ID, sampling start and end time, sampling mode identifier, data digest (lightweight hash), collection environment meta information (such as edge node location, network status, device health score), and local credibility score. The evidence header is stored along with each report and cache; Perform prior verification, including: data integrity check (simple hash check), format verification, time coherence verification (detect obvious time jumps or out-of-order sampling), and device state consistency detection (for example, if a sensor reports the same value for N consecutive samplings, it is marked as possibly stuck); For data that fails the verification, the edge node handles it according to the strategy (retry, local cache and mark, or trigger manual review); S23, local credibility scoring and lightweight anomaly detection on collected data, filtering noise or abnormal data, and triggering enhanced strategy to ensure key evidence integrity, specifically: The edge node maintains a local credibility score for each data source, which is based on device historical performance (fault rate, false alarm rate), calibration records, and stability of the last several samples. The score is reflected in the evidence header and is adjusted over time. Before the data enters the reporting channel, the edge node performs two types of lightweight intelligent judgment: Noise / stuck filter: use simple statistics or lightweight ML (Machine Learning) model (this embodiment uses an anomaly detector based on sliding window) to filter out sensor noise and sensor stuck data; Suspicious enhancement reminder: when detecting local anomalies but the edge confidence is insufficient (for example, a blurred GPS with slight drift appears in the image), the node will be labeled as "suspicious and suggest enhancement", and additional images will be taken according to the enhancement strategy in step S13 and more time series will be recorded for reporting; For data marked as high credibility but abnormal (for example, a high-credibility temperature control device reports a threshold value), the edge immediately triggers local buffer reporting and generates an emergency evidence header to ensure that key evidence is prioritized for processing; S24, divide data into segments according to event semantics, perform priority-aware compression, and use hierarchical caching strategy to ensure safe and complete reporting of key data when network is limited, specifically: The edge node divides the collected data into semantic segments according to event semantics and evidence header, and each segment contains data from at least one channel (for example, the packing segment contains key frames of the camera + temperature curve segment at that time + scan code record). The segmentation is performed according to the event correlation strategy in step S14, so that each segment can naturally serve as the smallest unit for subsequent fingerprint generation; For each segment, the edge applies intelligent compression strategy: For images, key frame extraction and differential frame storage (only significant change frames are retained); for time series data, key nodes (turning points / abnormal points) are retained + multi-resolution summary; for text and structured data, field-level summaries are made; The compression strategy is affected by the credibility score and the enhancement flag in the evidence header (high credibility + high priority segment uses lower loss compression or complete retention); When offline / disconnected, the edge maintains hierarchical caching: short-term cache is used for temporary fast reporting / backtracking (retaining complete data for the last several hours), and long-term cache is stored in compressed form with meta-information index for subsequent reporting; the cache strategy also records link state and expected reporting priority, and when the network is restored, it reports in priority batches; Each time the data packet is reported, it is accompanied by the corresponding local evidence header and a small index list.

[0020] S3, convert the edge collected data into multi-level and verifiable commodity fingerprints to realize the traceability management of the whole life cycle of the commodity; through the three-layer structure of L1 (edge summary), L2 (event-level fusion) and L3 (full-link fusion), combined with event semantics, credibility score and enhancement strategy, multi-modal feature fusion, abnormal marking and dynamic incremental updating are realized to ensure the uniqueness, interpretability and sustainable maintenance of the fingerprints, and to provide a reliable basis for federal notarization and full-link audit, specifically: S31, generate L1 fingerprint, generate edge summary fingerprint based on the semantic segmentation and local evidence header of step S2, including multi-modal feature summary, device information, time window and credibility score, to provide a basis for fast verification and subsequent fusion, specifically: Using each semantic segmentation and its local evidence header output by step S2, L1 fingerprint is generated; L1 fingerprint is a lightweight and quickly verifiable edge summary, which is used for subsequent rapid verification and cloud connection, and the fingerprint content includes but is not limited to: data summary (compressed data features processed by hash / summary algorithm), sampling mode, enhancement flag, local credibility score, sampling mode, enhancement flag, local credibility score; For multi-modal data (image, sensor curve, text record), channel separation processing + feature vector compression is adopted to generate summary, including but not limited to: For image data, key frame feature vector is extracted; For time series data, get the turning point sequence or feature curve vector; For text / structured data, get field summary + hash combination; L1 fingerprint records data integrity and time continuity indicators when it is generated, providing a trusted starting point for cloud L2 / L3 fusion; S32, generate L2 fingerprint, fuse all L1 fingerprints and context information within a single event, assign weights according to credibility and enhancement flag, and retain inter-event dependencies to provide interpretable fingerprints for event-level analysis and cross-event traceability, specifically: L2 fingerprint takes a single event as a unit, and fuses all L1 fingerprints and context information under the event to form an event-level fingerprint; the following factors are considered in the fusion process: Multi-channel weighted fusion: dynamically assign weights to each channel according to the edge credibility score of step S23 and the enhancement strategy marking of step S13; Time context integration: ensure the time sequence continuity of each channel data within the event, and correct possible small drifts; Inter-event dependency preservation: The pre-event influence recorded by the event correlation strategy in step S14 is used as the input of the subsequent L3 fusion; The fusion algorithm uses scalable vector-level processing to generate a vectorized L2 fingerprint, including: multi-modal feature vector, weight coefficient, anomaly label, and summary information; For events that trigger the enhanced strategy, the original enhanced data features are retained in the L2 fingerprint, and whether it is used for anomaly analysis and subsequent notarization is also recorded; S33, generate L3 fingerprint, integrate all event (L2 fingerprint) information of the same commodity, generate full-link commodity fingerprint, record event sequence, anomaly label and key features, realize cross-stage, cross-node and full-life cycle unique identification, specifically: L3 fingerprint integrates all event (L2 fingerprint) information of the same commodity to form a full-link commodity fingerprint; L3 fingerprint also reflects the full picture of the life cycle of the commodity, ensuring cross-stage, cross-node, and cross-scenario uniqueness and verifiability; The full-link fusion strategy includes but is not limited to: Event weight allocation, combined with step S13 micro-condition triggering, anomaly label, and step S23 trust score; Cross-event feature vector fusion: multi-level weighted average + key feature retention method is used; Anomaly information embedding: the full-link fingerprint records event anomalies, enhanced strategy triggering records, and paragraphs with low local trustworthiness; L3 fingerprint output includes: cross-event vectorized features, full-link event index and time sequence, anomaly and enhanced strategy labels, and verification summary (for quick integrity verification); S34, fingerprint verification and dynamic update, integrity and time continuity verification of L1-L3 fingerprint, combined with edge cache and enhanced strategy dynamic correction and incremental update, to ensure that the fingerprint is continuously reliable and verifiable during the life cycle, specifically: Each time L1-L3 fingerprint is generated, integrity and continuity verification is performed: hash digest consistency, time sequence continuity, and event dependency logic check; If minor anomalies (such as time drift, local data missing) are found, dynamic update is performed through the edge cache in step S24 and the enhanced strategy in step S13 to generate a corrected fingerprint; For new events or newly collected data in the life cycle of the commodity, only the relevant L2 / L3 fingerprint segment is updated, without the need to recalculate the full link, ensuring efficiency.

[0021] S4, distributed notarization and multi-party verification of the multi-layer fingerprints generated in step S3 through the federal anchor mechanism to realize the full-link traceability and tamper resistance of the commodity; the dynamic weighting, incremental updating and abnormal self-repairing strategies of node weight are adopted to ensure real-time and reliable verification and full-link integrity during the life cycle of the commodity, while supporting cross-organizational multi-party verification and audit, improving the reliability and interpretability of traceability management, specifically: S41, after the occurrence of a key event of the commodity, an anchor record is generated, including event ID, L2 / L3 fingerprint summary, timestamp, context label and edge node signature, a distributed ledger is formed through multi-node synchronization, and a multi-hash chain is used to associate the anchor with the previous event to realize full-link traceability and tamper resistance, specifically: After each key event (key node defined in step S1, enhanced event in step S24) in the life cycle of the commodity occurs, an anchor record is generated in the federal anchor network, which includes but is not limited to: event ID and L2 / L3 fingerprint summary, timestamp (such as the evidence header time in step S22), event context label (including abnormal label, enhanced strategy trigger flag), edge node signature or trusted identity certificate; The anchor record is synchronized through a distributed ledger or a multi-node storage system; each node can verify the received anchor to form a node consensus; And a multi-hash chain structure is adopted, each anchor not only records the event fingerprint, but also links to the previous anchor and related dependent events (event association defined in step S14), forming a traceable and tamper-proof chain; S42, after receiving the anchor, each node performs local verification, including fingerprint integrity, time continuity and evidence header consistency, and generates a node weight based on the verification result, a high weight (higher than a first threshold) represents complete consistency, a medium weight (between a second weight and the first weight, the first weight is greater than the second weight) represents a slight deviation, and a low weight (lower than the second weight) or a rejected record represents an anomaly, providing a dynamic weighting basis for subsequent anchor confirmation; S43, according to the node weight and verification result, the validity of the anchor is confirmed, the anchor points exceeding the threshold are recorded to the global link ledger, the dependency relationship of the previous events is updated to form a complete commodity chain, and an incremental updating mechanism is supported, when the commodity new event or fingerprint changes, only the related anchor and subsequent chain are updated, improving the notarization efficiency and real-time performance, specifically: According to the node verification result and weight, the anchor is confirmed: if the weight accumulation exceeds the set threshold, the anchor is confirmed to be valid; if the threshold is not reached, the anchor is pending data or triggers edge enhancement collection (step S24); The confirmed anchor is recorded to the global link ledger and the dependency relationship with the previous event is updated to form a complete traceable commodity chain; When new events occur in the product life cycle or the original event fingerprint is updated, only the affected anchor points and their subsequent chains are updated, without the need to reconstruct the entire chain, improving system efficiency; S44, multi-party verification allows regulatory agencies, supply chain parties or third-party audit nodes to independently verify anchors, check fingerprint consistency and link integrity, and abnormal handling mechanism triggers edge enhancement collection for slight deviation, generates alarm and joint review for serious abnormality, while supporting visual traceability query, realizing closed-loop self-repair and cross-organization trust management, specifically: Support multi-party verification, that is, regulatory agencies, supply chain participants, consumers or third-party audit nodes can access anchors and independently verify them, and the verification process includes checking the consistency of product event chains against L1 / L2 / L3 fingerprints and edge evidence headers, checking node weights and anchor confirmation records, and evaluating notarization credibility; Set up an abnormal handling mechanism. When local abnormalities (slight time drift or data missing) occur, trigger edge data backtracking and enhanced collection (step S24); when serious abnormalities (fingerprint mismatch or chain break) occur, generate an alarm, record an audit log, and trigger a joint review across nodes; Multiple parties can view the product's full life cycle event chain, combined with abnormal markers and enhanced records, to quickly locate problem nodes or links.

[0022] S5, based on the product full-link traceability query and abnormal tracking of step S3 fingerprint and step S4 federal anchor, supports multi-role access and dynamic incremental update, can visually display product life cycle events, detect abnormalities and locate responsible nodes, while providing real-time update and trusted audit functions, forming a closed-loop traceable management mechanism, specifically: S51, users query the full-link event chain through the unique identification of the product, the system integrates L1~L3 fingerprints, anchor notarization status, node weights and event context, generates a visual link diagram, displays time series, data channels and abnormal markers, realizes transparent and reliable product life cycle information query, specifically: Users (such as regulatory agencies or supply chain managers) input the unique identification of the product (L3 fingerprint ID or anchor ID), and the system realizes the acquisition of product full-link event data by querying anchor chains and L1~L3 fingerprints; The system integrates the following information: product full life cycle event sequence (based on event definition and dependency relationship of step S14), L2 fingerprint feature vector and abnormal / enhanced marker of each event, event collection node information, timestamp, credibility score, weight and notarization status of verified nodes (steps S42, S43); The query result can generate a graphical link display: horizontal display of time series events, vertical display of multi-channel data and credibility status, and abnormal event highlighting for quick understanding of product status; S52, the system performs multi-dimensional anomaly determination on the full-link event based on edge credibility, event sequence integrity, and node weight, automatically locates the abnormal event node, analyzes the abnormal type and influence range, and generates an abnormal report, provides accurate responsibility tracking and processing suggestions in combination with event context and fingerprint characteristics, specifically: The system performs multi-dimensional determination on the abnormality in the commodity event chain according to the query result: Low edge credibility: the score in step S23 is lower than the threshold; Event sequence anomaly: the event dependency relationship defined in step S14 is broken; Node verification anomaly: the node weight in step S42 is too low or the anchor point is not confirmed Define an abnormal tracking mechanism, including but not limited to: locating abnormal event nodes and influence range, automatically identifying abnormal types: data missing, time drift, fake or abnormal enhanced events, tracing event impact in combination with L1~L3 fingerprints, and outputting possible responsible nodes or links; The system can provide an abnormal report, including but not limited to: abnormal event sequence, credibility analysis, cross-node verification results, and recommended processing measures; S53, when a new event or an updated event is added in the commodity life cycle, the system only incrementally updates the related L1~L3 fingerprints and anchor chains, realizes real-time query and display, provides a new or modified event prompt, ensures that the query result is up-to-date, complete and efficient, without the need to recalculate the full link, and improves the system response capability and maintenance convenience, specifically: The system supports real-time query of new events or updated events in the commodity life cycle without the need to regenerate full-link fingerprints; Define an incremental update mechanism, including: automatically updating L3 fingerprints and related anchor chains when adding L1 / L2 fingerprints, locally refreshing the notarization status of updated events or abnormal events, and querying interfaces pulling the latest anchor chain and fingerprint information in real time to ensure data consistency, providing incremental update prompts to query users, and clearly indicating which events are new or modified to facilitate quick understanding of commodity state changes; S54, the system provides differentiated query permissions according to user roles (regulatory agencies, supply chain participants, and consumers), ensures data credibility in combination with anchor notarization, records access logs and abnormal viewing conditions, balances information transparency and security, realizes multi-party trusted query and audit protection, specifically: The system provides differentiated query and tracking permissions according to user roles: Regulatory agencies: can view full-link information and node weight for comprehensive audit of abnormalities; Supply chain participants: can view events and abnormalities related to their own links, and support self-checking and optimization; Consumer / third-party audit: accessible to the public part of the event and the full-link traceability result; Combined with the notarization result of the federal anchor point in step S4, it is ensured that the accessed data is a trusted and non-tamperable version; Support multi-dimensional logging, including access records, query operations and abnormal event viewing, to ensure traceability of the audit.

[0023] Embodiment two, as Figure 2 shown, the present application proposes a commodity traceability management system based on big data, which is used to execute the commodity traceability management method based on big data proposed in embodiment one, comprising: a multi-source data acquisition and preprocessing module, an edge evidence head generation and credibility enhancement module, a multi-layer fingerprint generation and incremental update module, a federal anchor point notarization and multi-party verification module, and a commodity traceability query and abnormal tracking module.

[0024] The multi-source data acquisition and preprocessing module is responsible for collecting original commodity data from each link of the supply chain, production equipment, logistics nodes and sales channels, including sensor data, event records, environmental information and operation logs, and simultaneously performing preliminary cleaning, formatting and time sequence correction on the data to generate uniform standardized event records, providing high-quality, structured multi-source data input for subsequent fingerprint generation and credible analysis; The edge evidence head generation and credibility enhancement module generates event evidence heads at the collection node end, records event time, collection device identity and preliminary credibility score, and combines local edge computing for data compression, summary and credibility enhancement, realizes event dynamic enhancement and abnormal marking through incremental evidence update strategy, ensures that each commodity event has verifiability and traceability at the source, and provides reliable input for fingerprint generation and chain notarization; The multi-layer fingerprint generation and incremental update module extracts features from the preprocessed multi-source data and edge evidence heads to generate L1, L2 and L3 three-layer fingerprints, which respectively represent basic event features, key node aggregation features and commodity life cycle summary fingerprints, and supports incremental update mechanism, when new events occur or fingerprints are updated in the commodity life cycle, only update related levels and subsequent dependent fingerprints, ensure system efficiency, continuity and scalability, at the same time provide event interdependence modeling, realize end-to-end full-link traceability; The federal anchor point notarization and multi-party verification module is responsible for distributing multi-layer fingerprints for anchor point notarization, synchronizing and confirming between federal nodes, each node verifying the anchor point and generating dynamic weight, confirming the effectiveness of the anchor point by integrating the opinions of multiple parties, and supporting incremental link maintenance, cross-node abnormal review and multi-party verification mechanism, realizing the notarization of the full-link of the commodity which is non-tamperable, at the same time, combining with the event context and node weight to ensure the credibility and explainability of the notarization result; The commodity traceability query and abnormality tracking module provides full-link commodity information query, visual display and multi-dimensional abnormality tracking functions, a user can query a full life cycle event chain according to a commodity identifier, the system realizes abnormality determination and responsibility positioning in combination with anchor point notarization and node weight, and supports dynamic incremental query and real-time update, can provide differentiated access permissions and trusted audit functions for supervisory agencies, supply chain participants and consumers, simultaneously generates an abnormality report and processing suggestions, and realizes closed-loop traceable management.

[0025] The embodiments of the present application are described in detail above in combination with the drawings, but the present application is not limited thereto, and various changes can be made within the knowledge of those skilled in the art without departing from the spirit of the present application.

Claims

1. A product traceability management method based on big data, characterized in that, The specific implementation steps include the following: S1. Construct a semantic event system for the product lifecycle, and clarify the data scope, triggering conditions, and related logic with previous and subsequent events for each type of event; S2. Collect, preprocess, and conduct preliminary credibility assessments of semantic events at edge nodes, and generate semantic segmentation data and evidence headers; S3. Transform edge-collected data into a multi-layered verifiable product fingerprint that combines L1 edge summarization, L2 event-level fusion, and L3 end-link fusion. S4. Distributed notarization and multi-party verification of the multi-level verifiable product fingerprints are performed through a federated anchor mechanism. S5. Conduct full-chain traceability and anomaly tracking of goods based on multi-level verifiable product fingerprints and federated anchor notarization results.

2. The product traceability management method based on big data according to claim 1, characterized in that, Building a semantic event system for the product lifecycle specifically includes: Based on product category, production process characteristics and regulatory requirements, a lifecycle semantic event framework is established, dividing the product lifecycle into several semantic segments with behavioral meaning and process constraints; Define the data types that can be collected, the data source devices, and the responsible parties for collecting data for each semantic event; Establish a micro-condition triggering mechanism for each type of event, and automatically trigger enhanced acquisition strategies based on sensor mutations, image anomalies, positional drift, or missing records. Establish the relationships between events and the fingerprint input strategy, and clarify that the data of the preceding events will be used as contextual information to participate in the fingerprint calculation of subsequent events.

3. The product traceability management method based on big data according to claim 2, characterized in that, Step S2 specifically includes: Automatically generate collection configurations based on event semantics and device capabilities, dynamically adjust sampling modes and frequencies, and realize adaptive access and contextualized sampling strategies for heterogeneous devices. Synchronize time and verify data integrity at edge nodes to generate an evidence header containing device information, sampling period, and local credibility. Perform local credibility scoring and lightweight anomaly detection on the collected data, filter out noise or anomalous data, and trigger enhancement strategies. Data is divided into semantic segments based on event semantics, priority-aware compression is performed, and a layered caching strategy is adopted to ensure the safe and complete reporting of critical data when the network is limited.

4. The product traceability management method based on big data according to claim 3, characterized in that, Step S3 specifically includes: L1 fingerprint is generated by generating an edge summary fingerprint based on semantic segmentation and local evidence header. The L1 fingerprint includes multimodal feature summary, device information, time window and confidence score. Generate L2 fingerprints, fuse all L1 fingerprints and context information within a single event, assign weights based on trustworthiness and enhancement flags, and preserve inter-event dependencies; Generate L3 fingerprints by integrating all event information of the same product to generate a full-chain product fingerprint, and record event sequences, anomaly markers and key features; Integrity and temporal continuity checks are performed on L1, L2, and L3 fingerprints, and dynamic corrections and incremental updates are performed using edge caching and enhancement strategies.

5. The product traceability management method based on big data according to claim 4, characterized in that, Generating an L1 fingerprint includes: Extracting keyframe feature vectors from image data; For time series data, obtain the sequence of inflection points or the vector of characteristic curves; For text or structured data, obtain field summaries and hash combinations; The L1 fingerprint records data integrity and temporal continuity indicators during generation.

6. The product traceability management method based on big data according to claim 5, characterized in that, Generating an L2 fingerprint includes: A multi-channel weighted fusion method is adopted, and the weights of each channel are dynamically allocated according to the edge confidence score and the enhancement strategy label. Integrate the time context to ensure the temporal continuity of data across channels within an event; Record the impact of preceding events according to the event association strategy.

7. The product traceability management method based on big data according to claim 6, characterized in that, Generating an L3 fingerprint includes: An event weighting strategy is adopted, combined with micro-condition triggering, anomaly marking, and credibility scoring; A multi-level weighted average and key feature preservation method is used for cross-event feature vector fusion; Embed anomaly information in the end-to-end fingerprint to record event anomalies, enhancement strategy trigger records, and segments with low local credibility.

8. The product traceability management method based on big data according to claim 7, characterized in that, Step S4 specifically includes: Anchor records are generated after key events of a product occur. These anchor records include an event ID, an L2 fingerprint digest or an L3 fingerprint digest, a timestamp, a context marker, and an edge node signature. A distributed ledger is formed through multi-node synchronization. After receiving the anchor point, each node performs local verification, including fingerprint integrity, temporal continuity and evidence header consistency checks, and generates node weights based on the verification results. The validity of anchor points is confirmed based on node weights and verification results. Anchor points exceeding the set threshold are recorded in the global link ledger, and the dependencies of preceding events are updated. It supports independent verification of anchor points by regulatory agencies, supply chain participants, or third-party audit nodes to check fingerprint consistency and chain integrity.

9. A product traceability management method based on big data according to claim 8, characterized in that, Step S5 includes: Users can query the entire event chain through the product's unique identifier. The system integrates L1 fingerprint, L2 fingerprint, L3 fingerprint, anchor point notarization status, node weight, and event context information to generate a visual chain diagram. The system performs multi-dimensional anomaly detection on end-to-end events based on edge credibility, event sequence integrity, and node weight, automatically locates abnormal event nodes, and analyzes the anomaly type and impact range. When a new event or an updated event is added during the product lifecycle, the system only incrementally updates the relevant L1 fingerprint, L2 fingerprint, L3 fingerprint, and anchor chain. The system provides differentiated query permissions based on user roles and combines anchor point notarization to ensure data credibility.

10. A big data-based commodity traceability management system, used to execute the big data-based commodity traceability management method according to any one of claims 1 to 9, characterized in that, include: The multi-source data acquisition and preprocessing module is used to collect raw commodity data from various links of the supply chain and perform preliminary cleaning, formatting and time-series correction to generate standardized event records. The edge evidence header generation and credibility enhancement module is used to generate event evidence headers at the acquisition node, record the event time, the identity of the acquisition device and the preliminary credibility score, and perform data compression, summarization and credibility enhancement. The multi-layer fingerprint generation and incremental update module is used to extract features from preprocessed multi-source data and edge evidence heads to generate L1 fingerprints, L2 fingerprints and L3 fingerprints, and supports an incremental update mechanism. The Federation Anchor Notarization and Multi-Party Verification module is used to perform distributed anchor notarization of multi-layer fingerprints, synchronize and confirm them among federation nodes, and support multi-party verification mechanisms. The product traceability and anomaly tracking module provides full-chain product information query, visualization, and multi-dimensional anomaly tracking functions.

Images