A multi-source perception constraint agricultural product circulation traceability method

Abstract

The present application relates to the technical field of agricultural product circulation traceability and quality safety risk control, and particularly relates to a multi-source perception constraint agricultural product circulation traceability method, which links agricultural product batches, shared media, events and evidence by unified dynamic identification relationship, forms shared medium state sequence and risk time window, constructs a dynamic exposure map, calculates risk inheritance non-dilution coefficients, and generates batch-level circulation control instructions and associated storage data by combining with lag detection evidence backtracking and recalculation. Through unified dynamic identification relationship, shared medium state sequence and dynamic exposure map, the shared medium risk time window and clean separation credibility are determined, the risk inheritance non-dilution coefficients are calculated, and based on the lag detection evidence backtracking and recalculation, the batch-level circulation control instructions and associated storage data are generated, so that the accuracy of risk identification, circulation control and storage review is improved.

Description

Technical Field

[0001] This invention relates to the field of agricultural product circulation traceability and quality and safety risk control technology, and more specifically, it relates to a multi-source sensing constraint method for agricultural product circulation traceability. Background Technology

[0002] Agricultural products are typically recorded and managed in batches throughout the harvesting, sorting, cleaning, temporary storage, transportation, processing, packaging, and sales processes. Existing traceability methods largely establish relationships between the origin, flow, test results, storage records, and sales destination of agricultural product batches, and have a solid foundation for routine quality traceability, flow tracking, and responsibility verification. However, with the increasing number of links in the agricultural product supply chain, faster flow speeds, and higher requirements for quality and safety management, relying solely on batch ledgers or static test records is no longer sufficient to fully cover complex situations such as multiple batches coming into contact with each other, reuse of media, and delayed arrival of test evidence.

[0003] In actual distribution, shared media such as turnover baskets, pallets, conveyor belts, washing tanks, refrigerated boxes, and transport vehicles often carry or come into contact with multiple batches of agricultural products at different times. If a batch of agricultural products has quality abnormalities, the risk may not only extend along the flow of that batch itself, but may also spread laterally between different batches through shared media. Furthermore, agricultural product batches may be split, merged, mixed, repackaged, or processed during the distribution process, further complicating the relationship between the original batch and the target batch. Current practices can usually record some cleaning, testing, or distribution information, but there is still a lack of unified dynamic correlation and calculation basis for determining the risk time window in which shared media exist, whether a cleaning event is sufficient to cut off the risk transmission relationship, and whether testing evidence should be retrospectively revised.

[0004] Furthermore, agricultural product testing evidence exhibits a certain time lag. For example, sampling may occur before the test report is issued, and by the time the test results arrive, the relevant agricultural product batch has already entered the storage, processing, distribution, or sales stages. If delayed testing evidence cannot be promptly traced back to the corresponding contact event or media cleaning event, it can easily lead to an overly narrow or overly broad assessment of the risk scope, thereby affecting the accuracy of circulation control measures such as suspending shipments, batch locking, retesting, and adjusting the recall scope.

[0005] Therefore, this application proposes a multi-source sensing constraint-based method for tracing the flow of agricultural products. Summary of the Invention

[0006] To address the shortcomings of existing technologies, the present invention aims to provide a method for tracing the flow of agricultural products with multi-source sensing constraints.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A multi-source sensing constraint method for agricultural product circulation traceability includes: executed by an agricultural product circulation traceability system, comprising the following steps: Step 1: Establish a unified and dynamic identification relationship between agricultural product batches, shared media, contact events, batch conversion events, media cleaning events, and testing evidence; Step 2: Collect multi-source sensing constraint data corresponding to agricultural product batches, shared media, contact events, batch conversion events, media cleaning events, and detection evidence; construct a shared media state sequence; and determine the shared media risk time window, cleaning isolation credibility, credible cleaning isolation events, and uncredible cleaning events based on the shared media state sequence. Step 3: Construct a dynamic exposure map based on unified dynamic identification relationships, shared media risk time windows, trusted cleaning isolation events, and untrusted cleaning events; Step 4: Calculate the risk inheritance non-dilution coefficient based on the dynamic exposure map, shared medium risk time window, cleaning isolation credibility, credible cleaning isolation events, and uncredible cleaning events, and determine the inheritance result of the restricted state based on the risk inheritance non-dilution coefficient. Step 5: Based on the delayed detection evidence, the dynamic exposure map, the risk time window of the shared medium, the confidence of the cleaning isolation and the non-dilution coefficient of risk inheritance are retrospectively recalculated, and batch-level circulation control instructions and associated evidence data of agricultural products are generated according to the results of the retrospective recalculation; among them, delayed detection evidence is the detection evidence whose arrival time is delayed compared to the corresponding contact event or the corresponding medium cleaning event.

[0008] In one embodiment, the contact event is an event generated when an agricultural product batch comes into contact with the shared medium, the batch conversion event is an event generated when an agricultural product batch is converted, the medium cleaning event is an event generated when the shared medium undergoes cleaning treatment, and the detection evidence is a data object corresponding to the detection results or complaint data related to the agricultural product batch or the shared medium.

[0009] In one embodiment, the unified dynamic identification relationship includes the association between agricultural product batch identification, shared medium identification, contact event identification, batch conversion event identification, medium cleaning event identification, detection evidence identification, and time information.

[0010] In one embodiment, the shared medium state sequence includes a medium contamination potential sub-state, a medium disinfection resilience sub-state, a medium load-bearing sub-state, a medium environmental stability sub-state, and a data reliability sub-state; the medium risk state of the shared medium in different time periods is determined based on the shared medium state sequence, and the continuous time period in which the medium risk state meets the preset risk conditions is determined as the shared medium risk time window.

[0011] In one embodiment, the reliability of the cleaning partition is determined based on the completeness of cleaning execution, the coverage of cleaning time sequence, the recovery degree of state before and after cleaning, the consistency of multi-source data, the support of detection evidence, and the integrity of the partition boundary; wherein, the completeness of cleaning execution is determined based on the execution type, execution duration, or execution object of the media cleaning event, the coverage of cleaning time sequence is determined based on the time coverage relationship between the media cleaning event and adjacent contact events, and the recovery degree of state before and after cleaning is determined based on the changes in the state of the shared media before and after the media cleaning event.

[0012] In one embodiment, a trusted cleaning interruption event is used to sever risk propagation relationships within a shared media risk time window, while an untrusted cleaning event is prohibited from severing risk propagation relationships within a shared media risk time window.

[0013] In one embodiment, the dynamic exposure map includes agricultural product batch nodes, shared media nodes, contact event nodes, batch transformation event nodes, media cleaning event nodes, and detection evidence nodes, as well as contact exposure edges, lateral propagation edges, transformation acceptance edges, cleaning gating edges, evidence mapping edges, and downstream flow edges; each node corresponds to an agricultural product batch, shared media, contact event, batch transformation event, media cleaning event, and detection evidence, respectively, and each edge is used to connect the corresponding node or the risk time window of the shared media.

[0014] In one embodiment, the risk inheritance non-dilution coefficient is a composite propagation constraint value used to determine whether the restriction state corresponding to the upstream agricultural product batch node continues to the target agricultural product batch node; in the dynamic exposure map, the agricultural product batch node with the determined risk level is used as the upstream agricultural product batch node, and the agricultural product batch node associated with the dynamic exposure map is used as the target agricultural product batch node.

[0015] In one embodiment, the inheritance result of the restriction status is determined as follows: when the risk inheritance non-dilution coefficient reaches the preset inheritance threshold, and the risk propagation relationship between the upstream agricultural product batch node and the target agricultural product batch node is not blocked by the trusted clean isolation event, the agricultural product batch corresponding to the target agricultural product batch node inherits the restriction status corresponding to the risk level of the upstream agricultural product batch node, and it is prohibited to directly reduce the restriction status based on the agricultural product weight ratio, agricultural product quantity ratio, batch mixing ratio or single negative test result.

[0016] In one embodiment, the batch-level circulation control instructions for agricultural products include at least one of the following: prohibiting batch mixing, prohibiting repackaging, suspending outbound shipments, switching to isolated processing lines, triggering re-testing, restricting entry into direct consumption channels, restricting entry into highly sensitive consumption channels, shortening sales time limits, batch locking, expanding the recall scope, or narrowing the recall scope.

[0017] Compared with the prior art, the present invention has the following beneficial effects: This invention constructs a shared medium state sequence by collecting multi-source sensing constraint data, and determines the shared medium risk time window, clean isolation credibility, credible clean isolation event and uncredible clean event accordingly. It can make detailed judgments on the state changes of shared medium in contact with different batches of agricultural products at different times, so that the contact exposure relationship, lateral propagation relationship and clean gating relationship in the dynamic exposure map have clear time boundaries and credible basis, which helps to reduce the uncertainty brought about by simply relying on clean records or static batch flow to judge the risk propagation range. By calculating the risk inheritance non-dilution coefficient under the joint constraints of dynamic exposure maps, shared medium risk time windows, clean isolation credibility, credible clean isolation events, and uncredible clean events, the inheritance result of the restricted state is determined accordingly. When lagging detection evidence arrives, the dynamic exposure maps, shared medium risk time windows, clean isolation credibility, and risk inheritance non-dilution coefficient are also recalculated retrospectively, thereby generating batch-level circulation control instructions and associated evidence data for agricultural products. This provides traceable and verifiable calculation basis for actions such as suspending outbound shipments, prohibiting repackaging, triggering re-testing, batch locking, expanding or narrowing the recall scope, and improving the accuracy, timeliness, and evidence integrity of agricultural product circulation risk control. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the overall process of a multi-source sensing constraint method for tracing the circulation of agricultural products according to the present invention. Figure 2 This is a schematic diagram illustrating the determination of the reliability of cleaning partitions and the distinction between reliable cleaning partition events and unreliable cleaning events in this invention. Figure 3 This is a schematic diagram of backtracking recalculation based on hysteresis detection evidence in this invention. Detailed Implementation

[0019] Reference Figures 1 to 3 A multi-source sensing constraint-based agricultural product circulation traceability method includes: executed by an agricultural product circulation traceability system, comprising the following steps: Step 1: Establish a unified and dynamic identification relationship between agricultural product batches, shared media, contact events, batch conversion events, media cleaning events, and testing evidence; incorporate the batch objects, media objects, event objects, and evidence objects that exist separately in the process of agricultural product circulation into the same identification system, so as to avoid the identification of agricultural product batches being fragmented due to different record sources in the stages of loading, transfer, temporary storage, processing, cleaning, and testing; By establishing a unified dynamic identification relationship between agricultural product batches, shared media, contact events, batch transformation events, media cleaning events, and testing evidence, the agricultural product circulation traceability system can identify the contact relationship between agricultural product batches and shared media, the continuity relationship before and after agricultural product batch transformation, the temporal relationship between media cleaning events and contact events, and the attribution relationship between testing evidence and corresponding agricultural product batches or shared media in the same data chain, providing a basic data structure for the construction of dynamic exposure maps and risk inheritance judgment.

[0020] When agricultural products enter the harvesting, grading, temporary storage, transfer, processing or sales circulation stages, an agricultural product batch identifier is generated for each batch of agricultural products, and a shared medium identifier is generated for carriers such as turnover baskets, pallets, conveyor belts, washing tanks, refrigerated boxes, and transport carriages that come into contact with multiple batches of agricultural products; when an agricultural product batch comes into contact with a shared medium, the contact start time, contact end time, contact location, contact method, contact area, load quantity and environmental conditions are recorded, and a contact event identifier is generated; When agricultural product batches are split, merged, mixed, repackaged, reprocessed, or transformed, the batch identifier before transformation, the batch identifier after transformation, the transformation time, the transformation ratio, the transformation location, and the transformation operation information are recorded, and a batch transformation event identifier is generated. When shared media undergo cleaning treatments such as rinsing, disinfection, drying, replacement of linings, or discontinuation and isolation, the cleaning start time, cleaning end time, cleaning method, cleaning intensity, personnel performing the cleaning, cleaning objects, and cleaning location are recorded, and a media cleaning event identifier is generated. When pesticide residue detection, microbial detection, foreign matter detection, abnormal temperature and humidity verification, consumer complaints, or regulatory spot checks are generated, the test results or complaint data are registered as test evidence, and a test evidence identifier is generated. Taking a batch of strawberries as an example, the agricultural product batch label corresponds to the strawberry batch with a picking date of June 3 and an origin of planting area A. The shared medium label corresponds to the turnover basket numbered A018. The contact event label corresponds to the contact record of the strawberry batch being loaded into turnover basket A018 between 8:20 and 9:10. The medium cleaning event label corresponds to the record of turnover basket A018 being disinfected at 10:30. The testing evidence label corresponds to the microbial test results formed by the regulatory sampling inspection on the same day. The system links contact events, batch conversion events, and testing evidence for the same agricultural product batch at different stages based on time information. It also links contact events and media cleaning events for the same shared medium during continuous use, intermittent cleaning, and re-containment. Testing evidence is then associated with the corresponding agricultural product batch identifier or shared medium identifier based on the sampling object, sampling location, sampling time, and testing object. When an agricultural product batch is repackaged or merged to form a new batch, the continuity between the pre-conversion and post-conversion batch identifiers is maintained to prevent the loss of the original source record after changes in name, packaging, or weight. When the testing evidence is formed later than the contact event or media cleaning event, it is still included in the original relationship chain according to the testing evidence identifier, time information, and corresponding object identifier. Through the above processing, a traceable dynamic association is formed between agricultural product batches, shared media, contact events, batch conversion events, media cleaning events, and testing evidence. For example, if turnover basket A018 carried another batch of blueberries before carrying this batch of strawberries, and there was a disinfection record between the two carrying events, if the disinfection record lacks an end time or the testing evidence indicates that turnover basket A018 has microbial abnormalities, the unified dynamic identification relationship can retain the complete association between the strawberry batch, the blueberry batch, turnover basket A018, the two contact events, the media cleaning event, and the corresponding testing evidence, providing a continuous and clear factual basis for risk time range identification and batch responsibility verification.

[0021] Step two involves collecting multi-source sensing constraint data corresponding to agricultural product batches, shared media, contact events, batch conversion events, media cleaning events, and testing evidence. This data forms a shared media state sequence, and based on this sequence, the risk time window, cleaning isolation credibility, credible cleaning isolation events, and uncredible cleaning events are determined. The multi-source sensing constraint data generated at the agricultural product circulation site is transformed into a continuous data foundation reflecting changes in the shared media's state. This multi-source sensing constraint data includes both agricultural product batch-related data and state data of the shared media during contact, exposure, cleaning, environmental changes, and testing feedback. The agricultural product circulation traceability system constructs a shared medium status sequence based on this data, enabling continuous recording of risk changes, cleaning effectiveness, and data reliability of the shared medium at different times. Determining the risk time window for the shared medium indicates the time range within which the shared medium is at high risk or where the risk has not been eliminated; determining the reliability of cleaning isolation allows for the distinction between whether a medium cleaning event is sufficient to prevent risk transmission; and the classification of reliable cleaning isolation events and unreliable cleaning events provides a basis for judging whether to retain, block, or restore risk transmission edges in the subsequent dynamic exposure map.

[0022] For shared media such as turnover baskets, pallets, washing tanks, sorting tables, conveyor belts, refrigerated boxes or transport carriages, the following data are collected for contact events: contact start time, contact end time, contact location, contact area, quantity carried, and batch of agricultural products in contact; batch conversion events: agricultural product batch before conversion, agricultural product batch after conversion, conversion ratio, and conversion time in batch conversion events; cleaning events: cleaning execution type, execution duration, execution object, cleaning agent concentration, cleaning temperature, cleaning coverage, and cleaning end time in media cleaning events; and sampling time, sampling location, testing object, testing results, and complaint content in the testing evidence. Same shared medium Multi-source sensing constraint data according to preset time granularity Write the state sequence. Pick to , Indicates a time index. Indicates the first Each point in time. Numerical raw indicators. Convert to basic normalized value using the following formula: Media contamination potential state and medium bearing load sub-state The risk is assigned a positive value, with higher values ​​indicating higher risk. The higher the value, the lower the risk. ; Media disinfection toughness substate Medium environment stability sub-state and data trust sub-state The index is selected based on the positive state, with larger values ​​indicating a better state. The larger the value, the worse the condition of the indicator. .in, Indicates the first The original index values ​​within a time granularity This indicates the lower limit of the equipment range, the lower limit of the process specification, or the verified historical sample number. percentile, This indicates the upper limit of the equipment's range, the upper limit of the process specification, or the first verified historical sample. Percentiles; when hour, Set to 0. When the weights in the same calculation formula are not specified by the enterprise's quality and safety specifications, regulatory sampling rules, or historical qualified sample sets, they shall be assigned equal weights. The historical qualified sample set is a shared set of media status samples that have been excluded from abnormal records and whose test results are qualified. When data related to cleaning effectiveness, test evidence support, and isolation boundary integrity is missing, the corresponding item shall be set to 0. When data related to media risk status is missing, the value of the most recent valid time granularity shall be used. When more than 3 consecutive time granularities are missing, the data credibility sub-state shall be set to 0.

[0023] Based on the aforementioned multi-source sensing constraint data, a shared media state sequence is formed in chronological order. The shared media state sequence includes the media contamination potential sub-state, the media disinfection resilience sub-state, the media load-bearing capacity sub-state, the media environmental stability sub-state, and the data reliability sub-state. Among them, the media contamination potential sub-state is determined by combining the risk level of the previous batch of agricultural products that came into contact with the shared media, the risk of residual substances on the surface of the shared media, and the degree of abnormality of the detection evidence; the media disinfection resilience sub-state is determined by combining the cleaning execution type, execution duration, cleaning agent concentration, and cleaning coverage; the media load-bearing capacity sub-state is determined by combining the number of batches of agricultural products that came into contact with the media per unit time, the weight carried, and the frequency of contact; the media environmental stability sub-state is determined by combining temperature, humidity, exposure duration, and cold chain fluctuations; and the data reliability sub-state is determined by combining the integrity of records, the continuity of timestamps, the stability of equipment acquisition, and the consistency of manual data entry. The shared medium state sequence is denoted as: in, Indicates the potential state of the medium contamination. Indicates the toughness state of the disinfected medium. Indicates the sub-state of the medium bearing the load. Represents the stability sub-state of the medium environment. This indicates the data trustworthiness sub-state. The result is obtained by weighted averaging based on the normalized values ​​of the respective data items; the number of data items for the same sub-state is... And when no weights are specified, each weight is taken as... The risk level of agricultural product batches is mapped to 0 for no abnormalities, 0.33 for moderate abnormalities, 0.67 for higher risk, and 1 for severe risk; when complaint data has not yet been verified, the mapping value is 0.5. (Testing evidence) When the result is qualitative, a negative result is taken. Positive results ; test evidence When quantifying results, detect the degree of abnormality. Determine using the following formula: in, Indicates the test evidence number, This represents the normalized value indicating the degree of anomaly detected. Indicates the detected value. This indicates the limit value, acceptance value, or internal control value of the corresponding testing item that is greater than 0; when there is no limit value for the testing item but there is an internal control value. Take the enterprise's internal control values. When Unable to obtain or In such cases, test evidence is treated as a qualitative result; if it cannot be converted into a qualitative result, Pick The test evidence was then marked as pending verification. The complaint data was verified to be true. Take 1; the complaint data was found to be inaccurate after verification. Take 0.

[0024] For example, if the shared medium numbered turnover basket A018 carries a batch of strawberries from 8:00 to 9:00, is rinsed and disinfected from 9:20 to 9:40, and carries a batch of blueberries from 10:00 to 11:00, and if the temperature and humidity records show a high humidity environment from 9:00 to 11:00, and the corresponding test evidence for the strawberry batch shows that the microbial levels are too high, then the medium contamination potential of turnover basket A018 increases and the medium environmental stability decreases during the corresponding time period. Read the shared media status sequence according to the preset time granularity, and check the media risk status. Determine using the following formula: in, These represent risk state weights, all of which are non-negative. Risk assessment variables Determined according to the following rules. When any risk condition is met: If none of the above conditions are met: in, Indicates the threshold of media risk status. Indicates the pollution potential threshold. Indicates the disinfection resilience threshold. Indicates the load-bearing threshold. Indicates the environmental stability threshold. This represents the data reliability threshold. When there is no historical qualified sample set, Take 0.60, Take 0.70, Take 0.40, Take 0.70, Take 0.40, Take 0.50; when a historical qualified sample set exists, Take the corresponding variable percentile, Take the corresponding variable Percentiles. If to Continuously established, and or As the starting point of the sequence, or If the sequence ends, the shared medium risk time window is denoted as: in, and Indicates a time index. Indicates the risk time window number.

[0025] The reliability of the cleaning partition is determined based on the completeness of cleaning execution, the coverage of cleaning time sequence, the degree of state recovery before and after cleaning, the consistency of multi-source data, the support of detection evidence, and the integrity of the partition boundary. Among them, the completeness of cleaning execution is determined based on the execution type, execution duration, or execution object of the media cleaning event; the coverage of cleaning time sequence is determined based on the time coverage relationship between the media cleaning event and adjacent contact events; and the degree of state recovery before and after cleaning is determined based on the changes in the state of the shared media before and after the media cleaning event. Media cleaning event Reliability of cleaning partitions Determine using the following formula: in, Indicates the completeness of cleaning execution. Indicates cleaning sequence coverage. Indicates the degree of recovery before and after cleaning. Indicates consistency of multi-source data. Indicates the degree of support of the test evidence. Indicates the integrity of the partition boundary. to All are non-negative values, and . Divide the number of items meeting the standards by five factors: cleaning type, cleaning duration, cleaning agent concentration, cleaning objects, and cleaning coverage. Values ​​are assigned as follows: 1 for items that meet the criteria, and 0 for items that do not meet the criteria or are missing. Assume a media cleaning event. The start time is The end time is The interval of the previous contact event is The interval of the next contact event is Cleaning sequence coverage Determine using the following formula: when hour, Pick When there is no prior contact event, and The intersection length is counted as 0; when there is no subsequent contact event, and The intersection length is counted as 0. The coverage risk boundary refers to a media cleaning event located between two adjacent contact events that establish a lateral propagation relationship, where the cleaning end time is later than the end time of the previous contact event and earlier than the start time of the subsequent contact event. Reaching the isolation boundary threshold .

[0026] set up This represents the average media risk state within the time granularity preceding the start of cleaning. The media risk state is the average value within a certain time granularity after cleaning, representing the degree of state recovery before and after cleaning. Determine using the following formula: The ratio of the number of consistent data items among equipment records, manual records, cleaning logs, timestamps, and testing evidence to the total number of data items involved in the verification is calculated. The total number of data items involved in the verification is... hour, Take 0. Take the sampling coverage after cleaning and The product of This represents the normalized value indicating the degree of abnormality detected after cleaning; when no evidence of abnormality was obtained after cleaning, Take 0. Where, The ratio of the area of ​​the sampled area after cleaning to the actual area of ​​the cleaned area is taken, with an upper limit of 1; if the actual area of ​​the cleaned area is 0 or cannot be determined, the ratio is taken. Set to 0. Boundary integrity of the partition. Determine using the following formula: in, Indicates the actual area cleaned. This represents the union of the contact areas of adjacent contact events before and after a media cleaning event; When empty, Set to 0. When a media cleaning event lacks records of cleaning start time, cleaning end time, cleaning object, or actual cleaning coverage area, If all values ​​are 0, the media cleaning event is recorded as an untrusted cleaning event. Cleaning isolation trust threshold. Pick , isolation boundary threshold Set the threshold to 0.80 to support the evidence. Take 0.60; when When a media cleaning event has a cleaning end time and a record of the object being cleaned, the media cleaning event... This is recorded as a trusted cleaning isolation event; if any condition is not met, it is a media cleaning event. This is recorded as an untrusted cleaning event.

[0027] When the credibility of the cleaning partition reaches the preset credibility threshold, and the media cleaning event covers the risk boundary of the shared media risk time window, the media cleaning event is identified as a credible cleaning partition event. The credible cleaning partition event cuts off the risk propagation relationship within the shared media risk time window. When the credibility of the cleaning partition is lower than the preset credibility threshold, or the cleaning execution record lacks a cleaning end time, the cleaning object does not cover the actual contact area, or the detection evidence still shows that there is an anomaly in the shared media, the media cleaning event is identified as an uncredible cleaning event. Uncredible cleaning events are prohibited from cutting off the risk propagation relationship within the shared media risk time window. For example, if the turnover basket A018 only recorded manual rinsing from 9:20 to 9:40 without recording the disinfectant concentration, and the swab test evidence formed at 10:15 showed that the microorganisms were still positive, then the integrity of the cleaning execution and the support of the test evidence for this media cleaning event are both insufficient, the credibility of the cleaning isolation is lower than the preset credibility threshold, and this media cleaning event is determined to be an uncredible cleaning event. The risk transmission relationship between the strawberry batch and the subsequent blueberry batch is still retained.

[0028] Step 3: Based on unified dynamic identification relationships, shared medium risk time windows, credible cleaning isolation events, and uncredible cleaning events, a dynamic exposure map is constructed. The relationships between agricultural product batches, shared media, contact events, batch conversion events, media cleaning events, and testing evidence are organized from a single record format into a graphical exposure relationship structure. When constructing the dynamic exposure map, the agricultural product circulation traceability system uses unified dynamic identification relationships as the basis for object matching, the shared medium risk time window as the time boundary for risk propagation, and credible cleaning isolation events and uncredible cleaning events as gating conditions for whether risk propagation has been interrupted. Through dynamic exposure mapping, agricultural product batches are no longer tracked solely according to logistics or processing order. Instead, they can be combined with shared media contact, batch transformation, cleanliness isolation, and testing evidence mapping to present the exposure path and associated range of agricultural product batches in the real circulation environment.

[0029] Agricultural product batch nodes are generated according to agricultural product batch identifiers, shared media nodes are generated according to shared media identifiers, contact event nodes are generated according to contact event identifiers, batch conversion event nodes are generated according to batch conversion event identifiers, media cleaning event nodes are generated according to media cleaning event identifiers, and detection evidence nodes are generated according to detection evidence identifiers. The time information, location record, object identifier, and event attributes of each node are written into the corresponding node attributes. Among them, the agricultural product batch node records the batch source, quantity, weight, circulation location and risk level; the shared medium node records the medium type, medium number, historical batches carried and status change records; the contact event node records the start and end time, contact method and contact range of the contact between the agricultural product batch and the shared medium; the batch conversion event node records the inheritance relationship between the agricultural product batch before conversion and the agricultural product batch after conversion; the medium cleaning event node records the cleaning method, cleaning duration, execution object and cleaning isolation credibility; and the detection evidence node records the sampling object, sampling time, detection results or complaint data. Taking the shared medium numbered A018 as an example, the strawberry batch node, blueberry batch node, A018 node, two contact event nodes, one medium cleaning event node, and one microbial detection evidence node are included in the same dynamic exposure map. Corresponding identifiers and time information are retained between each node to avoid the source break or media reuse relationship loss after batch conversion when relying solely on manual ledgers. Based on the node set, contact exposure edges, lateral propagation edges, transformation acceptance edges, cleanliness gating edges, evidence mapping edges, and downstream flow edges are established. Specifically: when agricultural product batch nodes and shared medium nodes have a temporal overlap relationship under the same contact event node, a contact exposure edge is established between the agricultural product batch node, the contact event node, and the shared medium node; when two or more agricultural product batch nodes successively contact the same shared medium node within the same shared medium risk time window, and there is no credible cleanliness isolation event between them, a lateral propagation edge is established between the relevant agricultural product batch nodes; when agricultural product batches undergo splitting, merging, mixing, repackaging, repackaging, or processing transformation, a transformation acceptance edge is established between the agricultural product batch node before transformation, the batch transformation event node, and the agricultural product batch node after transformation. When a media cleaning event node is located between adjacent contact event nodes, a cleaning gating edge is established according to the credibility of the cleaning isolation. The cleaning gating edge corresponding to a credible cleaning isolation event blocks the risk propagation relationship within the shared media risk time window, while the cleaning gating edge corresponding to an uncredible cleaning event retains the risk propagation relationship within the shared media risk time window. When a detection evidence node has a corresponding relationship with an agricultural product batch node or a shared media node in terms of sampling object, sampling location, sampling time, or complaint object, an evidence mapping edge is established. When an agricultural product batch node flows into the warehousing, processing, distribution, sales, or recall process, a downstream flow edge is established. Taking turnover basket A018 as an example, if the strawberry batch comes into contact with turnover basket A018 between 8:00 and 9:00, and the blueberry batch comes into contact with the same turnover basket A018 between 10:00 and 11:00, and there is a cleaning record between 9:20 and 9:40, but this cleaning event is determined to be an unreliable cleaning event, then the dynamic exposure map retains a lateral propagation edge between the strawberry batch node and the blueberry batch node, and connects the microbial detection evidence node formed at 10:15 to the turnover basket A018 node through the evidence mapping edge, forming a complete exposure path that includes contact, cleaning, detection and downstream circulation relationships.

[0030] Step four: Based on the dynamic exposure map, the risk time window of the shared medium, the credibility of the cleaning isolation, the credible cleaning isolation event and the uncredible cleaning event, calculate the risk inheritance non-dilution coefficient, and determine the inheritance result of the restriction status according to the risk inheritance non-dilution coefficient; on the basis of the already formed dynamic exposure map, constrain the risk inheritance relationship between agricultural product batches, rather than simply reducing the restriction status according to the weight ratio, quantity ratio or batch mixing ratio of agricultural products. The agricultural product traceability system combines the associated paths in the dynamic exposure map, the risk persistence within the shared medium risk time window, the reliability of clean isolation, the blocking effect of reliable clean isolation events, and the preserving effect of unreliable clean events on risk propagation relationships to calculate the risk inheritance non-dilution coefficient. This risk inheritance non-dilution coefficient participates in determining the inheritance result of the restriction status, enabling a traceable and verifiable judgment of the continuation of the restriction status between agricultural product batches with determined risk levels and those associated with the dynamic exposure map. This reduces the possibility of prematurely lifting the risk status due to mixing, repackaging, transshipment, or a single negative test result.

[0031] In the dynamic exposure map, agricultural product batch nodes with determined risk levels are identified as upstream agricultural product batch nodes, and agricultural product batch nodes associated through contact exposure edges, lateral propagation edges, transformation acceptance edges, cleanliness gating edges, evidence mapping edges, or downstream circulation edges are identified as target agricultural product batch nodes. For each association path between upstream agricultural product batch nodes and target agricultural product batch nodes, the risk time window of the shared medium traversed by the association path, the duration of the contact event, the contact range, the batch conversion ratio, the shared medium carrying capacity, the detection evidence results, and the credibility of the cleanliness barrier are read. The exposure path strength, risk time overlap, cleanliness barrier reverse constraint value, unreliable cleanliness retention value, detection evidence support value, and transformation acceptance retention value are set. Set upstream agricultural product batch nodes To the target agricultural product batch node exist The associated path. For the first... Related paths ,like There exist trusted cleaning isolation events that cover the risk boundary, and trusted cleaning blocking variables. ;otherwise . No. Path inheritance value of associated paths Determined according to the following rules. When a reliable clean-blocking mechanism exists: When a reliable clean blockade is not available: in, Indicates the intensity of the exposed path. Indicates the degree of overlap in risk time. This indicates the reverse constraint value of the cleaning partition. This indicates an unreliable cleanliness retention value. This indicates the value supporting the test evidence. Indicates the conversion and retention value. to All are non-negative values, and .

[0032] For path Contact events within Normalized value of contact duration Pick , Indicates the start time of contact. Indicates the end time of contact. Take the allowable contact duration or the duration of a historical qualified contact event for this type of shared media process. Percentiles; Normalized value of contact area Pick , Indicates the actual contact area. Indicates the accessible area of ​​the shared medium; risk value of the contact method. Calculate 1 for direct exposed contact, 0.5 for liner contact, 0.1 for sealed packaging contact, and 0 for no contact. Path In Take each contact event within the path. The arithmetic mean; when there are no contact events within the path. All values ​​are set to 0. Exposure path intensity. Determine using the following formula: set up Representing a path The set of contact events in Indicates contact event The union of all risk time windows of the shared media; if no risk time window exists, It is an empty set. Risk time overlap. Determine using the following formula: when Empty, or path The sum of the durations of all contact events is less than or equal to hour, Take 0. If the path The credibility of the cleaning partition exists. Below The media cleaning event, Pick When no such media cleaning event occurs, Take 1. If If there are untrusted cleaning events, then ;otherwise Detection evidence support value Take the evidence set detected along this path The maximum value; when the evidence set is empty, . Sampling time should fall within the corresponding contact interval or shared medium risk time window. The interval between the above intervals shall not exceed Pick In other cases, take . Take the sampling coverage area and path The ratio of medium-risk contact area to area, with an upper limit of 1; the area of ​​high-risk contact area is... Or when it cannot be determined, Set to 0. Risk exposure areas are determined by the path. Contact locations and contact areas that overlap in the risk time window of the internal and shared media are determined by combining them; when there is regional overlap, the overlapping part is counted only once. If If it includes a lateral propagation edge or a contact exposed edge, then ;like Only through the conversion of the receiving edge connection, then Pick , This represents the product of the weight or quantity proportions of each batch of conversion events entering the target agricultural product batch. Take 0.60; if the conversion ratio record is missing, Take 1.

[0033] to Risk inheritance non-dilution coefficient Determine using the following formula: Preset inheritance threshold Take 0.60. When and and When the risk propagation relationship between them is not blocked by a credible cleaning isolation event, inherit The risk level corresponds to the restriction status. In a single negative test... Setting it to 0 only creates a detection record and does not represent a reduction. Deductions; when , Number of re-inspections for the same object and the same risk item At the same time, upon establishment, this negative test result is registered as representative negative evidence, forming a verification marker, and does not directly reduce [the effectiveness / level]. .

[0034] For example, if the strawberry batch node has been identified as a high-risk level, and the strawberry batch came into contact with turnover basket A018 between 8:00 and 9:00, and the blueberry batch came into contact with the same turnover basket A018 between 10:00 and 11:00, and turnover basket A018 is in the shared media risk time window between 9:00 and 11:00, and the media cleaning event between 9:20 and 9:40 is identified as an unreliable cleaning event due to the lack of disinfectant concentration records and abnormal swab test results, then the lateral propagation edge between the strawberry batch node and the blueberry batch node still participates in the risk inheritance non-dilution coefficient calculation; The inheritance result of the restriction status is determined based on the risk inheritance non-dilution coefficient. Specifically, the calculated risk inheritance non-dilution coefficient is compared with a preset inheritance threshold, and it is verified whether the risk propagation relationship between the upstream agricultural product batch node and the target agricultural product batch node is blocked by a trusted clean isolation event. When the risk inheritance non-dilution coefficient reaches the preset inheritance threshold, and the risk propagation relationship is not blocked by a trusted clean isolation event, the restriction status corresponding to the risk level of the upstream agricultural product batch node is determined for the agricultural product batch corresponding to the target agricultural product batch node. This restriction status includes suspension of outbound shipment, restriction of circulation, batch locking, expanded testing, or entry into isolation disposal. In the process of determining the inheritance result, the agricultural product weight ratio, agricultural product quantity ratio, batch mixing ratio, and single negative test result are only used as auxiliary records and are not used as the basis for directly reducing the restriction status. Taking a batch of blueberries as an example, even if the blueberry batch only comes into contact with the turnover basket A018 for 30 minutes, and the subsequent single blueberry sample tested negative, as long as the risk inheritance non-dilution coefficient between the strawberry batch node and the blueberry batch node reaches the preset inheritance threshold, and there is no credible cleaning isolation event between the two, the blueberry batch will still inherit the restriction status under the corresponding risk level of the strawberry batch, thus avoiding premature lifting of control due to a small mixing ratio, low weight ratio, or a single negative result.

[0035] Step 5 involves retrospectively recalculating the dynamic exposure map, shared medium risk time window, cleanliness isolation confidence, and risk inheritance non-dilution coefficient based on lagging detection evidence. The results of this retrospective recalculation are then used to generate batch-level circulation control instructions for agricultural products and associated evidence data. Lagging detection evidence refers to evidence whose arrival time lags behind the corresponding contact event or corresponding medium cleaning event. This addresses the problem of existing traceability results being difficult to correct in a timely manner when the arrival time of detection evidence is later than the on-site contact event or medium cleaning event. In the process of agricultural product circulation, there is often a time lag between sampling, submission for testing, issuance of results, and entry into the system. If the original dynamic exposure map, shared medium risk time window, cleanliness barrier reliability, and risk inheritance non-dilution coefficient are still used as fixed results, it is easy to miss agricultural product batches that have already been circulated, transformed, or come into contact with shared media. The agricultural product circulation traceability system performs retrospective recalculation based on lagging test evidence, which can correct the originally determined risk time range, cleanliness barrier reliability, and restriction state inheritance results, and generate agricultural product batch-level circulation control instructions and associated evidence data based on the retrospective recalculation results. The agricultural product batch-level circulation control instructions form control actions for specific agricultural product batches, while the associated evidence data retains the basis for retrospective recalculation, map version, evidence association, and status change records, which facilitates regulatory verification, responsibility tracing, and subsequent handling review.

[0036] When the arrival time of the test evidence is later than the corresponding contact event or the corresponding media cleaning event, the test evidence is registered as delayed test evidence, and its test object, sampling time, sampling location, evidence arrival time, test result, complaint data source, and corresponding agricultural product batch identifier or shared media identifier are read; the test evidence node, agricultural product batch node, shared media node, contact event node, and media cleaning event node in the dynamic exposure map are repositioned based on the delayed test evidence, and the start and end boundaries of the shared media risk time window are corrected by the sampling time rather than the evidence arrival time; Delayed detection evidence The sampling time is recorded as The time of arrival of the evidence is recorded as The evidence object node is denoted as The backtracking time interval is denoted as: In the atlas version In the middle, read the valid time interval and Intersect and with The set of nodes with a connected path is denoted as the affected node set. Read the valid time interval and Intersecting and both ends of the node belong Let the set of affected edges be denoted as the set of edges affected by the edge. When delayed detection evidence Evidence object node When it is a shared medium node, Take Contains The set. When lagging detection evidence Evidence object node When dealing with batch nodes for agricultural products, read and... A set of shared media that exists within the sampling time or the most recent circulation period, including contact exposure edges, conversion receiving edges, or downstream circulation edges. ; Each shared medium in the process is corrected for the shared medium risk time window. When empty, only the risk level of the agricultural product batch node and the mapping edge of the detection evidence are updated; the risk time window of the shared medium is not corrected. (Later detection evidence) Display shared media exist When an anomaly exists, The corresponding time granularity is denoted as the risk time granularity; the backtracking starting point is taken as... The end time of the most recent trusted cleaning isolation event is taken as the starting point of the shared medium state sequence when no trusted cleaning isolation event exists; the end time of the backtracking is taken as... Then, the start time of the most recent trusted cleaning isolation event is used; if no trusted cleaning isolation event occurs, the end time of the shared medium state sequence is used. Between the backtracking start point and the backtracking end point... The continuous time granularity is merged into a corrected shared medium risk time window. Window difference set This includes adding risk time windows, deleting risk time windows, and risk time windows whose start and end times have changed.

[0037] If the delayed detection evidence shows that the shared medium still has anomalies after a certain medium cleaning event, the cleaning execution integrity, cleaning time sequence coverage, state recovery before and after cleaning, multi-source data consistency, detection evidence support, and isolation boundary integrity of the medium cleaning event are recalculated, and the cleaning isolation credibility is corrected. When the corrected cleaning isolation credibility is lower than the preset credibility threshold, the originally identified credible cleaning isolation event is adjusted to an uncredible cleaning event, the risk propagation relationship that was cut off within the risk time window of the shared medium is restored, and the risk inheritance non-dilution coefficient between the relevant upstream agricultural product batch node and the target agricultural product batch node is recalculated. Taking turnover basket A018 as an example, if the shared medium has a medium cleaning event between 9:20 and 9:40, and carries a batch of blueberries between 10:00 and 11:00, and the swab test evidence arriving at 11:30 shows that there is microbial abnormality at the sampling location at 9:50, then the test evidence is used as lagging test evidence for retrospective recalculation, and the time point at 9:50 is included in the risk time window of the shared medium. The medium cleaning event between 9:20 and 9:40 is adjusted to an unreliable cleaning event due to insufficient support from the test evidence. After retrospectively recalculating the dynamic exposure map, shared medium risk time window, clean isolation credibility, and risk inheritance non-dilution coefficient, the affected agricultural product batch nodes, downstream circulation edges, transformation and acceptance edges, and horizontal propagation edges are verified one by one. Based on whether the risk inheritance non-dilution coefficient reaches the preset inheritance threshold, whether the risk propagation relationship is blocked by the credible clean isolation event, whether the agricultural product batch has been repackaged or repackaged, and whether the agricultural product batch has entered the sales or processing stage, corresponding agricultural product batch-level circulation control instructions are generated. Agricultural product batch-level circulation control instructions include at least one of the following: prohibiting batch mixing, prohibiting repackaging, suspending outbound shipment, switching to the isolation processing line, triggering re-testing, restricting entry into direct consumption channels, restricting entry into highly sensitive consumption channels, shortening the sales time limit, batch locking, expanding the recall scope, or shrinking the recall scope. For affected agricultural product batch nodes If backtracking and recalculation And upstream risk agricultural product batch nodes With regard to the affected agricultural product batch nodes If the risk propagation relationship between them is not blocked by a trusted cleaning isolation event, then instructions to suspend outbound shipments, lock batches, trigger re-inspection, or restrict flow are generated; if If the product has already entered the downstream distribution, sales, or processing stages, an order to expand the recall scope will be generated. If Between all upstream high-risk agricultural product batch nodes All below If the corresponding risk propagation relationships are all blocked by a credible clean isolation event, a review release or recall scope reduction instruction will be generated. This is in response to conflicting test evidence and issues with sampling coverage. Below Or data trust sub-state Below At this time, only a command to trigger a re-test is generated; no command to cancel the review or narrow the recall scope is generated. Conflict of test evidence refers to the same test object, the same test item, and time intervals not exceeding [a certain threshold]. The test evidence contains both positive and negative results, or the difference in the degree of abnormality in the quantitative test evidence is greater than [missing value]. The set of affected nodes is increased by linking the evidence storage data. Set of affected edges Window Difference Set Atlas version number Instruction generation time and backtracking recalculation rule number ,in, Indicates upstream risk agricultural product batch node To the affected agricultural product batch nodes The risk inheritance non-dilution coefficient, This indicates the generation time of batch-level circulation control instructions for agricultural products. This indicates the rule number that triggered this backtracking recalculation.

[0038] For example, if a blueberry batch has moved from the warehouse to the sorting stage but has not yet entered the sales end, and the retrospective recalculation results show that there is a risk propagation relationship between it and the strawberry batch that was not blocked by a credible clean isolation event, and the risk inheritance non-dilution coefficient reaches a preset inheritance threshold, then an agricultural product batch-level circulation control instruction will be generated for that blueberry batch, including suspending outbound shipment, prohibiting repackaging, triggering re-inspection, and switching to an isolated processing line. If some downstream batches have already been sold and distributed, the recall scope will be expanded based on the downstream circulation edge. The associated evidence records include the delayed detection evidence identifier, evidence arrival time, sampling time, shared medium risk time window before and after retrospective recalculation, changes in clean isolation credibility, changes in the risk inheritance non-dilution coefficient, the content of the agricultural product batch-level circulation control instruction, and the generation time, forming a verifiable chain of evidence.

[0039] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for tracing the flow of agricultural products with multi-source sensing constraints, characterized in that, include: The process, executed by the agricultural product circulation traceability system, includes the following steps: Step 1: Establish a unified and dynamic identification relationship between agricultural product batches, shared media, contact events, batch conversion events, media cleaning events, and testing evidence; Step 2: Collect multi-source sensing constraint data corresponding to agricultural product batches, shared media, contact events, batch conversion events, media cleaning events, and detection evidence; construct a shared media state sequence; and determine the shared media risk time window, cleaning isolation credibility, credible cleaning isolation events, and uncredible cleaning events based on the shared media state sequence. Step 3: Construct a dynamic exposure map based on unified dynamic identification relationships, shared media risk time windows, trusted cleaning isolation events, and untrusted cleaning events; Step 4: Calculate the risk inheritance non-dilution coefficient based on the dynamic exposure map, shared medium risk time window, cleaning isolation credibility, credible cleaning isolation events, and uncredible cleaning events, and determine the inheritance result of the restricted state based on the risk inheritance non-dilution coefficient. Step 5: Based on the delayed detection evidence, the dynamic exposure map, the risk time window of the shared medium, the confidence of the cleaning isolation and the non-dilution coefficient of risk inheritance are retrospectively recalculated, and batch-level circulation control instructions and associated evidence data of agricultural products are generated according to the results of the retrospective recalculation; among them, delayed detection evidence is the detection evidence whose arrival time is delayed compared to the corresponding contact event or the corresponding medium cleaning event.

2. The method for tracing the flow of agricultural products with multi-source sensing constraints according to claim 1, characterized in that, Contact events are events generated when agricultural product batches come into contact with shared media; batch conversion events are events generated when agricultural product batches are converted; media cleaning events are events generated when shared media undergo cleaning treatment; and detection evidence is the data object corresponding to the detection results or complaint data related to agricultural product batches or shared media.

3. The method for tracing the flow of agricultural products with multi-source sensing constraints according to claim 2, characterized in that, The unified dynamic identification relationship includes the association between agricultural product batch identification, shared medium identification, contact event identification, batch conversion event identification, medium cleaning event identification, testing evidence identification, and time information.

4. The method for tracing the flow of agricultural products with multi-source sensing constraints according to claim 3, characterized in that, The shared medium state sequence includes the medium contamination potential sub-state, the medium disinfection toughness sub-state, the medium load-bearing sub-state, the medium environmental stability sub-state, and the data reliability sub-state. Based on the shared medium state sequence, the medium risk state of the shared medium in different time periods is determined, and the continuous time period in which the medium risk state meets the preset risk conditions is determined as the shared medium risk time window.

5. The method for tracing the flow of agricultural products with multi-source sensing constraints according to claim 4, characterized in that, The reliability of the cleaning partition is determined based on the completeness of cleaning execution, the coverage of cleaning time sequence, the degree of state recovery before and after cleaning, the consistency of multi-source data, the support of detection evidence, and the integrity of the partition boundary. Among them, the completeness of cleaning execution is determined based on the execution type, execution duration, or execution object of the media cleaning event; the coverage of cleaning time sequence is determined based on the time coverage relationship between the media cleaning event and adjacent contact events; and the degree of state recovery before and after cleaning is determined based on the changes in the state of the shared media before and after the media cleaning event.

6. The method for tracing the flow of agricultural products with multi-source sensing constraints according to claim 5, characterized in that, Trusted cleaning isolation events are used to cut off risk propagation relationships within the risk time window of shared media, while untrusted cleaning events are prohibited from cutting off risk propagation relationships within the risk time window of shared media.

7. The method for tracing the flow of agricultural products with multi-source sensing constraints according to claim 1, characterized in that, The dynamic exposure map includes agricultural product batch nodes, shared media nodes, contact event nodes, batch transformation event nodes, media cleaning event nodes, and detection evidence nodes, as well as contact exposure edges, lateral propagation edges, transformation acceptance edges, cleaning gating edges, evidence mapping edges, and downstream flow edges. Each node corresponds to an agricultural product batch, shared media, contact event, batch transformation event, media cleaning event, and detection evidence, respectively, and each edge is used to connect the corresponding node or the risk time window of the shared media.

8. The method for tracing the flow of agricultural products with multi-source sensing constraints according to claim 7, characterized in that, The risk inheritance non-dilution coefficient is a composite propagation constraint value used to determine whether the restriction status corresponding to the upstream agricultural product batch node continues to the target agricultural product batch node. In the dynamic exposure map, the agricultural product batch node with the determined risk level is used as the upstream agricultural product batch node, and the agricultural product batch node associated with the dynamic exposure map is used as the target agricultural product batch node.

9. A method for tracing the flow of agricultural products with multi-source sensing constraints according to any one of claims 1-8, characterized in that, The inheritance result of the restriction status is determined as follows: when the risk inheritance non-dilution coefficient reaches the preset inheritance threshold, and the risk propagation relationship between the upstream agricultural product batch node and the target agricultural product batch node is not blocked by the trusted clean isolation event, the agricultural product batch corresponding to the target agricultural product batch node inherits the restriction status corresponding to the risk level of the upstream agricultural product batch node, and it is prohibited to directly reduce the restriction status based on the agricultural product weight ratio, agricultural product quantity ratio, batch mixing ratio or single negative test result.

10. The method for tracing the flow of agricultural products with multi-source sensing constraints according to claim 9, characterized in that, The batch-level circulation control instructions for agricultural products include at least one of the following: prohibiting batch mixing, prohibiting repackaging, suspending outbound shipments, switching to isolated processing lines, triggering re-testing, restricting entry into direct consumption channels, restricting entry into highly sensitive consumption channels, shortening sales time limits, batch locking, expanding the recall scope, or narrowing the recall scope.

Images