Rfid electronic tag data traceability system and method

Abstract

The application relates to the field of wireless radio frequency identification and Internet of Things tracing technology, in particular to an RFID electronic tag data tracing system and method; the system comprises a data acquisition module, a benchmark reconstruction module, a variation simulation module, a residual error generation module, a coupling judgment module and a tracing feedback module; electronic tag digital identification data, underlying radio frequency characteristic data and tracing node space-time data are acquired, an ideal benchmark characteristic vector is reconstructed, an abnormal state is injected to generate a damaged simulation characteristic vector, actual measurement residual error and simulation residual error are formed and are subjected to time alignment and scale normalization, similarity calculation and environment coupling scoring are combined to output physical tampering, environmental noise or to-be-recovered judgment, and tracing chain blocking, radio frequency compensation or reacquisition instructions are generated; the application can distinguish between environmental fluctuations and malicious tampering, reduce false positives and false negatives, and improve the credibility of the tracing chain.

Description

Technical Field

[0001] This invention relates to the fields of radio frequency identification and Internet of Things (IoT) traceability technology, specifically to an RFID electronic tag data traceability system and method. Background Technology

[0002] Currently, RFID electronic tag data traceability is usually achieved based on the correspondence between the electronic tag's digital identifier and the business records of the traceability node. However, when electronic tags are subjected to environmental interference such as metal reflection, liquid absorption, or personnel obstruction during cold chain circulation, warehousing and handling, or node acceptance, or when abnormal situations such as tag peeling and re-attaching, or duplication and forgery occur, the existing traceability methods cannot effectively distinguish between environmental fluctuations and physical tampering, which can easily lead to misjudgments or omissions, reducing the accuracy and reliability of electronic tag traceability results. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention provides an RFID electronic tag data traceability system and method. Specifically, the technical solution of this invention is as follows: The RFID electronic tag data traceability system includes a data acquisition module, a benchmark reconstruction module, a variation simulation module, a residual generation module, a coupling decision module, and a traceability feedback module. The data acquisition module is used to acquire the electronic tag digital identification data containing the tag identification code, the underlying radio frequency feature data, and the spatiotemporal data of the traceability node collected by the RFID reader, so as to generate a multi-dimensional acquisition vector on site. The benchmark reconstruction module is used to call the preset ideal radio frequency digital twin model, and based on the tag identification code and traceability node spatiotemporal data in the field multidimensional acquisition vector, to perform benchmark reconstruction on the field multidimensional acquisition vector to generate an ideal benchmark feature vector; The mutation simulation module is used to inject physical mutations into the ideal baseline feature vector based on a preset abnormal state simulation model to generate a damaged simulation feature vector. The residual generation module is used to generate measured residual feature sets and simulated residual feature sets based on field multi-dimensional acquisition vectors, ideal benchmark feature vectors and damaged simulation feature vectors, and to perform time alignment and scale normalization processing to generate comparable residual feature sets. The coupling decision module is used to calculate the similarity of comparable residual feature sets to generate a tampering coupling score. It calculates the measured residual feature set based on a preset feature template to generate an environmental coupling score, and combines the environmental coupling score to generate a physical tampering decision result, an environmental noise decision result, or a decision result to be re-sampled. The traceability feedback module is used to generate a traceability chain blocking instruction in response to a physical tampering judgment result, to generate an RF compensation instruction in response to an environmental noise judgment result and send it to the RFID reader, or to generate a re-collection instruction in response to a judgment result to be re-collected.

[0004] Preferably, the data acquisition module includes a digital identifier acquisition unit, a radio frequency feature acquisition unit, and a node spatiotemporal acquisition unit; The digital identification acquisition unit is used to obtain electronic product codes, tag identification codes, and encrypted anti-counterfeiting verification codes from electronic tags in order to generate electronic tag digital identification data; The radio frequency feature acquisition unit is used to obtain the received signal strength indication time series, radio frequency phase angle and antenna backscattering cross section parameters from the RFID reader to generate the underlying radio frequency feature data; The node spatiotemporal acquisition unit is used to obtain the spatial coordinates and timestamps of the traceable nodes in order to generate spatiotemporal data of the traceable nodes.

[0005] Preferably, the baseline reconstruction module includes a model parameter calling unit and an ideal state generation unit; the model parameter calling unit is used to call the tag identification code, the preset ideal dielectric constant of the tag backplane, and the preset free space loss model in the electronic tag digital identification data; The ideal state generation unit is used to generate an ideal reference feature vector based on the tag identification code, the spatiotemporal data of the traceability node, the ideal dielectric constant of the tag backplane, and the free space loss model. The ideal reference eigenvector includes the ideal received signal strength attenuation curve and the expected value of the ideal RF phase angle.

[0006] Preferably, the mutation simulation module includes a structural damage mutation unit, a cloning mutation unit, and a simulation state output unit; The structural damage variation unit is used to call the preset antenna micro-deformation impedance mutation operator and the preset backplane dielectric constant step factor, and inject them into the corresponding dimension in the ideal reference feature vector to change the expected value of the ideal radio frequency phase angle and the ideal received signal strength indication attenuation curve. The clone mutation unit is used to call the preset non-original chip capacitance tolerance factor and inject it into the ideal reference feature vector to change the resonant characteristics of the tag chip; The simulation output unit is used to generate a damaged simulation feature vector based on the injection results of the structural damage mutation unit and the clonal mutation unit.

[0007] Preferably, the residual generation module includes a measured residual extraction unit, a simulation residual extraction unit, and a synchronous normalization unit; The measured residual extraction unit is used to perform differential processing between the field multidimensional acquisition vector and the ideal benchmark feature vector to generate a measured residual feature set; The simulation residual extraction unit is used to perform differential processing between the damaged simulation feature vector and the ideal baseline feature vector to generate a simulation residual feature set; The synchronization normalization unit is used to perform time alignment and scale normalization on the measured residual feature set and the simulated residual feature set to generate a comparable residual feature set.

[0008] Preferably, the coupling decision module includes a similarity calculation unit, an environment coupling degree calculation unit, and a state determination unit; The similarity calculation unit is used to calculate the comparable residual feature set based on the multidimensional dynamic time warping algorithm or the cosine similarity algorithm to generate the tampering coupling score; The environmental coupling calculation unit is used to calculate the measured residual feature set based on the preset multipath fading feature template and the preset occlusion attenuation feature template to generate an environmental coupling score. The state determination unit is used to compare the tampering coupling score with a preset tampering threshold and the environmental coupling score with a preset environmental threshold. When the tampering coupling score is greater than or equal to the tampering threshold, the physical tampering judgment result is output. When the tampering coupling score is less than the tampering threshold and the environmental coupling score is greater than or equal to the environmental threshold, the environmental noise decision result is output. When the tampering coupling score is less than the tampering threshold and the environmental coupling score is less than the environmental threshold, the decision result for re-collection is output.

[0009] Preferably, the traceability feedback module includes a link control unit, a compensation control unit, and a re-collection control unit; The link control unit is used to respond to the physical tampering judgment result, generate a traceability chain blocking instruction to prevent the traceability record corresponding to the electronic tag from being written to subsequent traceability nodes, and write the electronic tag digital identification data, underlying radio frequency feature data and traceability node spatiotemporal data into the abnormal traceability record; The compensation control unit is used to generate, in response to the environmental noise decision result, an RF compensation command based on the environmental coupling score for adjusting at least one of the RFID reader's transmit power, operating frequency, receive gain, or sampling time window, and sends the RF compensation command to the RFID reader; The re-collection control unit is used to generate a re-collection command in response to the re-collection decision result and mark the current status of the traceability node as pending confirmation.

[0010] The RFID electronic tag data traceability method includes the following steps: Acquire digital identification data of electronic tags, underlying radio frequency feature data, and spatiotemporal data of traceability nodes collected by RFID readers to generate on-site multidimensional acquisition vectors; The preset ideal radio frequency digital twin model is invoked, and the tag identification code and traceability node spatiotemporal data in the field multidimensional acquisition vector are used to perform benchmark reconstruction of the field multidimensional acquisition vector in order to generate an ideal benchmark feature vector. Based on a pre-defined abnormal state simulation model, physical mutations are injected into the ideal baseline feature vector to generate a damaged simulation feature vector. Based on the on-site multidimensional acquisition vector, ideal benchmark feature vector, and damaged simulation feature vector, the measured residual feature set and the simulation residual feature set are generated, and time alignment and scale normalization are performed to generate a comparable residual feature set. The comparable residual feature set is calculated, and the measured residual feature set is calculated based on the preset feature template to generate an environmental coupling score. The environmental coupling score is then combined to generate a traceability validity judgment result. Based on the traceability validity judgment, a traceability chain blocking instruction, an RF compensation instruction, or a re-acquisition instruction is generated.

[0011] Preferably, physical mutation injection includes: The preset antenna micro-deformation impedance mutation operator and the preset backplate dielectric constant step factor are invoked, and the antenna micro-deformation impedance mutation operator and the backplate dielectric constant step factor are injected into the ideal reference feature vector to generate structural damage simulation features. Call the preset non-original chip capacitor tolerance factor and inject the non-original chip capacitor tolerance factor into the ideal benchmark feature vector to generate cloned damaged simulation features; Based on the structural damage simulation features and the clone damage simulation features, a damage simulation feature vector is generated.

[0012] Preferably, the traceability validity judgment and feedback control include: calculating the comparable residual feature set based on the multidimensional dynamic time warping algorithm or the cosine similarity algorithm to generate the tampering coupling score; Based on the preset multipath fading feature template and the preset occlusion attenuation feature template, the measured residual feature set is calculated to generate an environmental coupling score. When the tampering coupling score is greater than or equal to the preset tampering threshold, a physical tampering judgment result and a traceability chain blocking instruction are generated. When the tampering coupling score is less than the tampering threshold and the environmental coupling score is greater than or equal to the preset environmental threshold, an environmental noise decision result and an RF compensation command are generated. When the tampering coupling score is less than the tampering threshold and the environmental coupling score is less than the environmental threshold, a decision result for re-collection and a re-collection instruction are generated.

[0013] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention establishes a multi-dimensional acquisition vector for the same reading session by synchronously collecting and binding electronic tag digital identification data, underlying radio frequency feature data, and spatiotemporal data of traceability nodes. This enables a unified representation of tag identity information, radio frequency physical status, and node contextual information. It can solve the problem that existing traceability methods rely solely on the correspondence between digital identification and business records, making it difficult to reflect the true physical status of the tag. This improves the integrity and accuracy of the traceability data foundation. 2. This invention generates an ideal baseline feature vector by calling an ideal radio frequency digital twin model and combining the tag's chip identity and the spatiotemporal conditions of the traceability node. This establishes an individualized reference framework for the tag under undamaged and untampered conditions, avoiding errors caused by using historical averages as comparison objects and enhancing the physical targeting of subsequent anomaly identification. By injecting structural damage mutations and clonal mutations into the ideal baseline feature vector, a damaged simulation feature vector is constructed. This transforms anomaly risks such as tearing and re-labeling, chip replacement, or copying and forgery into a calculable and comparable physical anomaly model, thereby providing a simulateable and traceable basis for judging tampering behavior. 3. This invention generates a set of measured residual features and a set of simulated residual features, and performs time alignment and scale normalization to form a set of comparable residual features. This effectively eliminates irrelevant influences caused by differences in reading distance, dwell time, sampling rhythm and node equipment, allowing the system to focus on the residual morphology and structure itself, thus improving the comparison stability in complex flow scenarios. 4. This invention establishes a dual-scoring decision mechanism that combines tampering coupling analysis and environmental coupling analysis in parallel by calculating the similarity of comparable residuals and generating an environmental coupling score by combining multipath fading feature templates and occlusion attenuation feature templates. This mechanism can distinguish between environmental fluctuations and physical tampering, reducing the misjudgment of environmental interferences such as metal reflection, liquid absorption, and personnel occlusion as tampering, while also reducing the probability of missing anomalies such as peeling and re-attaching, cloning and forgery. 5. This invention generates traceability chain blocking instructions, radio frequency compensation instructions, and re-acquisition instructions based on three types of judgment results: physical tampering, environmental noise, and re-acquisition. This forms a closed-loop control mechanism from acquisition, reconstruction, simulation, comparison, judgment to feedback processing. It can not only block the writing of subsequent traceability records in a timely manner when an anomaly is confirmed and preserve the abnormal evidence chain, but also automatically adjust the reader's transmission power, operating frequency, receiving gain, or sampling window under environmental noise conditions, thereby improving the continuous operation capability and business adaptability of the field traceability system. 6. By setting fault-tolerant control mechanisms such as missing test markers, low weight markers, pending confirmation status, and manual review nodes, this invention can maintain cautious output under boundary conditions such as short-term offline status of the reader / writer, clock drift, local frost, sampling interruption, or missing parameters, avoiding direct conclusions of tampering when evidence is insufficient, and further improving the accuracy, credibility, and end-to-end application security of the traceability results. Attached Figure Description

[0014] The present invention will be further explained below with reference to the accompanying drawings and embodiments: Figure 1 This is a schematic diagram of the modules of the RFID electronic tag data traceability system provided in the embodiments of this application; Figure 2 This is a flowchart illustrating the RFID electronic tag data traceability method provided in the embodiments of this application. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0016] The RFID electronic tag data traceability system includes a data acquisition module, a benchmark reconstruction module, a variation simulation module, a residual generation module, a coupling decision module, and a traceability feedback module. The data acquisition module is used to acquire the electronic tag digital identification data containing the tag identification code, the underlying radio frequency feature data, and the spatiotemporal data of the traceability node collected by the RFID reader, so as to generate a multi-dimensional acquisition vector on site. The benchmark reconstruction module is used to call the preset ideal radio frequency digital twin model, and perform benchmark reconstruction based on the tag identification code and traceability node spatiotemporal data in the field multidimensional acquisition vector to generate an ideal benchmark feature vector; The mutation simulation module is used to inject physical mutations into the ideal baseline feature vector based on a preset abnormal state simulation model to generate a damaged simulation feature vector. The residual generation module is used to generate measured residual feature sets and simulated residual feature sets based on field multi-dimensional acquisition vectors, ideal benchmark feature vectors and damaged simulation feature vectors, and to perform time alignment and scale normalization processing to generate comparable residual feature sets. The coupling decision module is used to calculate the similarity of comparable residual feature sets to generate a tampering coupling score. It calculates the measured residual feature set based on a preset feature template to generate an environmental coupling score, and combines the environmental coupling score to generate a physical tampering decision result, an environmental noise decision result, or a decision result to be re-sampled. The traceability feedback module is used to generate a traceability chain blocking instruction in response to a physical tampering judgment result, to generate an RF compensation instruction in response to an environmental noise judgment result and send it to the RFID reader, or to generate a re-collection instruction in response to a judgment result to be re-collected.

[0017] This embodiment provides a mechanism for an RFID electronic tag data traceability system, such as... Figure 1 As shown; specifically, the entire traceability scenario of human cold chain vaccines, from production and packaging, regional warehousing, trunk transportation to hospital pharmacy acceptance, is the main focus; each box of vaccines is affixed with an ultra-high frequency RFID electronic tag, and the tag establishes a unique traceability relationship with the box of vaccines after filling; since vaccines are repeatedly transferred in different places such as cold storage shelves, insulated turnover boxes, metal pallets, and hospital receiving windows, the tag reading and writing may be affected by environmental factors such as metal reflection, liquid absorption, and personnel obstruction, and may also encounter the risk of the tag being torn off and transferred to other medicine boxes, or being counterfeited by copying chips; in order to avoid misjudging environmental fluctuations as tampering, or misjudging counterfeiting events as normal events, this embodiment adopts a continuous mechanism of ideal benchmark reconstruction, physical variation simulation, dual-track residual comparison, coupled decision and feedback processing; The data acquisition module reads the tag once or multiple times for a short period at each traceability node, forming a multi-dimensional acquisition vector on site. The multi-dimensionality refers not only to the encoded content, but also to the physical state presented during the radio frequency propagation process. Under the original cardboard material, standard attachment posture, and conventional reading distance, the echo intensity change, phase change, and backscatter state of a legitimate tag will present a relatively stable physical profile. However, once the tag is peeled off, bent, or transferred to a different material surface, the electrical length of the antenna, the impedance matching state, and the backplane medium conditions will change synchronously, leaving observable traces in the radio frequency characteristics received by the reader. The benchmark reconstruction module does not simply use historical averages as the comparison object, but calls a preset ideal radio frequency digital twin model. Based on the tag identification code and the spatiotemporal conditions of the current node, it reconstructs the ideal benchmark feature vector that the tag should exhibit under the conditions of being undamaged, untampered, and in an idealized propagation environment. By introducing an ideal benchmark corresponding to the tag identity, node position, and reading time, the ideal signal profile can be separated from the field observation. The mutation simulation module injects physical mutations into the ideal reference feature vector based on the abnormal state simulation model. Tearing off the reattachment will cause local deformation of the antenna, adhesive residue and changes in the attached substrate, which will cause characteristic distortion of the relationship between phase trajectory and intensity attenuation. Although the cloned label may replicate the surface markings, the chip parasitic parameters and resonant points are often not completely consistent with the original chip. After injecting these mechanistic offsets into the ideal reference, the simulation module obtains the damaged simulation feature vector. The residual generation module forms two comparison paths: one is to differiate the field multidimensional acquisition vector with the ideal reference feature vector to obtain the measured residual feature set; the other is to differiate the damaged simulation feature vector with the ideal reference feature vector to obtain the simulation residual feature set; time alignment and scale normalization are performed to avoid misjudgment caused by differences in relative speed, angle and dwell time when the tag passes through the antenna beam. The coupling decision module performs similarity calculations on comparable residual feature sets and analyzes whether the measured residuals conform to environmental templates such as multipath fading or occlusion attenuation to obtain an environmental coupling score. If the field residuals and the tampered simulation residuals closely match in terms of change profiles, it indicates that the offset source is closer to the physical state of the tag has changed. If the field offset is more consistent with the propagation disturbance caused by shelf reflection, condensation blockage, or stacking of turnover boxes, it is determined to be environmental noise. If neither type of feature is sufficient, it enters the waiting-for-re-sampling state. Furthermore, in this embodiment, the actual algorithm input of the coupled decision module is the comparable residual feature vector generated based on the comparable residual feature set of the same reading session; the aforementioned similarity calculation on the comparable residual feature set is used to explain the range of data sources on which the decision is based, and the similarity calculation on the comparable residual feature vector is used to explain the specific input form when the algorithm is executed. The two correspond to the same batch of residual data and do not constitute a change in the processing object. The traceability feedback module performs differentiated handling based on different judgment results; for physical tampering, the system generates a traceability chain blocking instruction to prevent the vaccine box from being written to subsequent nodes and saves the abnormal record; for environmental noise, the system does not block the business, but issues a radio frequency compensation instruction to adjust the reading parameters and then tries to collect data stably again; for data to be collected again, a re-collection instruction is issued and the current node status is marked as pending confirmation. Furthermore, in this embodiment, the comparable residual feature set refers to the set of multi-dimensional residual sequences such as received signal strength indication residual, phase residual, and scattering residual obtained after time alignment and scale normalization within the same reading session; the comparable residual feature vector refers to a single comparison input extracted and organized from the comparable residual feature set according to a preset dimensional order, time order, and missing measurement marking rules. Specifically, the former emphasizes the organizational form of residual data, while the latter emphasizes the input form when it is sent to similarity calculation. During system operation, the residual generation module first outputs a comparable residual feature set, and the coupled decision module then generates a corresponding comparable residual feature vector based on the comparable residual feature set within the same session to perform similarity calculation. Therefore, the two are not different technical objects, but rather the corresponding expressions of the same batch of residual data at different processing stages. Furthermore, the generation order of the aforementioned comparable residual feature vectors remains consistent across all nodes, prioritizing the arrangement according to the dimensions of received signal strength indication, phase, and scattering, and then further arranging them according to the session timestamp within each dimension. If a dimension is missing, a missing marker or a low-weight marker is retained at the corresponding position without changing the arrangement order of the other dimensions. With the help of this fixed rule, the comparison inputs generated by different nodes, different sessions, and different anomaly templates remain isomorphic, thereby ensuring that the similarity calculation in the coupled decision module is consistent with the calculation of the environment template. In actual deployment, if situations such as short-term offline of the reader, clock drift of the temperature and humidity gateway, or local frost on the tag causing instantaneous reading breaks occur, the system will prioritize retaining the original collection segments and node logs, and will not directly output the tampering conclusion. Instead, it will first enter the waiting re-collection process. If multiple re-collections still cannot form an effective comparable residual, the node can be marked as a manual verification node, and the operator can use a close-range handheld device or unpacking and sampling inspection to supplement the confirmation. When a batch of vaccines was transferred from the regional disease control warehouse to the hospital pharmacy for acceptance, the outer layer of the medicine box had an aluminum foil insulation lining. When the reader first read the box, the received signal strength of multiple tags indicated that the attenuation was greater than the preset attenuation threshold. The offset shape of most of the tags was consistent with the overall attenuation caused by the proximity of metal. The system judged it as environmental noise and automatically lowered the working frequency and extended the sampling window before rereading. The traceability chain continued normally. However, although the digital identification of the label of one box of vaccines was still readable, the phase offset showed a local abrupt change compared with the ideal state. It was also highly consistent with the preset simulated residual of tearing and re-adhesion. The system output a physical tampering judgment and blocked the registration of that box of vaccines in the hospital's warehouse. The purpose of this embodiment is to unify the digital identity verification of the label with the radio frequency physical status verification into the same traceability closed loop, so as to distinguish and deal with environmental fluctuations and malicious tampering, reduce false alarms and missed alarms, and improve the credibility of the drug traceability chain.

[0018] Furthermore, the data acquisition module includes a digital identification acquisition unit, a radio frequency feature acquisition unit, and a node spatiotemporal acquisition unit. The digital identification acquisition unit is used to obtain the electronic product code, tag identification code, and encrypted anti-counterfeiting verification code from the electronic tag to generate electronic tag digital identification data. The radio frequency feature acquisition unit is used to obtain the received signal strength indication time series, radio frequency phase angle, and antenna backscatter cross section parameters from the RFID reader to generate underlying radio frequency feature data. The node spatiotemporal acquisition unit is used to obtain the spatial coordinates and timestamps of the traceability node to generate traceability node spatiotemporal data.

[0019] This embodiment provides a refined mechanism for the data acquisition module. In the vaccine cold chain traceability scenario, if only the label code is collected without the underlying radio frequency characteristics, the system can only confirm the read number information, but cannot determine whether the label corresponding to the number still maintains its original physical form. Conversely, if only the radio frequency waveform is viewed without identity data and node spatiotemporal information, it is difficult to correlate the signal offset with a specific medicine box or a specific node. Therefore, this embodiment divides the acquisition process into three parts: digital identification acquisition, radio frequency characteristic acquisition, and node spatiotemporal acquisition, and completes the binding within the same reading session. The digital identification acquisition unit is used to acquire electronic product codes, label identification codes, and encrypted anti-counterfeiting verification codes. The electronic product code is used to identify the vaccine product and packaging level, making it easy to correspond with batch numbers, specifications, and flow records in the business system. The label identification code is used to characterize the unique physical identity of the label chip itself, which is usually solidified during the chip manufacturing stage and is difficult to completely rewrite. The encrypted anti-counterfeiting verification code is used to enhance the legality verification of the digital layer and prevent simple copying based solely on public codes. The combination of the three forms multi-layered digital identification data of business identity, chip identity, and security verification. The radio frequency (RF) feature acquisition unit is used to acquire the received signal strength indication time series, RF phase angle, and antenna backscattering cross section parameters. The received signal strength indication time series reflects the process of the tag's echo strength changing with its relative position within a continuous sampling window. The RF phase angle is sensitive to changes in antenna electrical length and propagation path, making it suitable for identifying microscopic changes after the tag is damaged, stretched, or reattached. The backscattering cross section parameters reflect the tag's comprehensive response capability to incident electromagnetic waves and can supplement information on scattering patterns in addition to intensity and phase. Furthermore, in engineering implementation, the antenna backscattering cross section parameter can be the scattering-related characteristics directly provided by the RFID reader, or it can be the equivalent scattering response parameter calculated from the amplitude and phase sampling values, the received signal strength indication time series, and the received link gain parameters returned by the reader after completing the reader power calibration, antenna gain calibration, and node field calibration. Specifically, this embodiment does not limit the reader hardware to natively output an independent register data called backscatter cross section. As long as the equivalent parameter that monotonically corresponds to the tag's scattering capability can be stably obtained in the same session, it can be used as part of the underlying radio frequency feature data. To ensure comparability between different nodes, the system prioritizes using the same type of reader and the same calibration process to output this parameter. If the node device models are different, the dimensional mapping is first completed on the node side, and then written into the field multidimensional acquisition vector. The node spatiotemporal acquisition unit is used to record the spatial coordinates and timestamps of the current traceability node; the spatial coordinates are used to distinguish different read and write environments such as the production line packing position, cold storage shelf position, and hospital receiving station; the timestamps are used to characterize time-related radio frequency environment changes such as day-night temperature difference, warehouse door opening and closing, and peak handover times. In a single reading session, the system can generate, for example, a digital portion consisting of electronic product code 1, tag identification code 1, and verification code 1; a radio frequency portion consisting of intensity segment A, phase segment A, and scattering segment A; and a spatiotemporal portion consisting of node coordinates P1 and time T1. The subsequent system processes a set of bound identity, physical, and spatiotemporal three-dimensional data. Furthermore, as an anomaly handling mechanism, if the digital identifier is read completely but only a portion of the radio frequency feature is collected, the system retains the collected fields and marks them as incomplete samples, and prioritizes triggering supplementary sampling in the future; if the radio frequency feature is complete but the verification code fails to be read, the system prohibits direct entry into the release process and requires a second card reading at close range; if the node coordinates are missing or the timestamp is abnormal, the on-site data is only temporarily stored and does not enter the baseline reconstruction process. Furthermore, if the backscattering cross section parameter is temporarily unavailable due to node calibration failure, the system does not directly regard its absence as a label anomaly, but instead records the dimension as a missing dimension, and reduces the weight of the dimension in subsequent residual comparisons or requires recalibration and re-sampling, so as to avoid misjudging the equipment-side inaccuracy as label tampering. At the hospital pharmacy receiving point, the operator used a fixed gantry to read the entire box of vaccines. The electronic product code and verification code of a certain box of vaccines were returned normally. However, because the angle between the edge of the box and the gantry antenna exceeded the preset angle threshold, only a received signal strength indication segment with a length lower than the preset segment threshold was obtained on the first attempt. The system marked it as incomplete sampling and required the box posture to be adjusted for a second reading. When the length of the received signal strength indicator segment is greater than or equal to the preset segment threshold, the system determines that the sampling is complete and triggers the subsequent reference reconstruction and residual generation process; another box of vaccines has complete radio frequency characteristics, but the node gateway has not yet synchronized to the pharmacy coordinates. The system waits for the location data to be completed before performing subsequent processing. The purpose of this embodiment is to provide complete and one-to-one corresponding identity, physical, and spatiotemporal basic data for subsequent ideal benchmark reconstruction and anomaly detection.

[0020] Furthermore, the benchmark reconstruction module includes a model parameter calling unit and an ideal state generation unit; the model parameter calling unit is used to call the tag identification code, the preset ideal dielectric constant of the tag backplane, and the preset free space loss model from the electronic tag digital identification data; the ideal state generation unit is used to generate an ideal benchmark feature vector based on the tag identification code, the spatiotemporal data of the tracing node, the ideal dielectric constant of the tag backplane, and the free space loss model; the ideal benchmark feature vector includes the ideal received signal strength indication attenuation curve and the expected value of the ideal radio frequency phase angle.

[0021] This embodiment provides a refinement mechanism for the benchmark reconstruction module. If only the average intensity or average phase read in the past is used as the reference benchmark, once a batch of vaccines changes its packaging material, the reading angle deviates, or the tag chip batches are different, the average value may lose its physical meaning and misjudge normal individual differences as abnormalities. Therefore, this embodiment introduces a model parameter calling unit and an ideal state generation unit to reconstruct the performance that the tag should have in the ideal state based on the tag identity and node environmental conditions. The model parameter calling unit reads the tag identification code of the tag; the tag identification code is not only an identification mark, but can also be mapped to the corresponding chip model, antenna matching parameter range, and factory calibration information; for the same vaccine manufacturer, the tags of different supply batches may have differences in process details, and these differences should not be mistaken for signs of tampering; in addition, this unit retrieves the ideal dielectric constant of the tag back plate, which can be understood as the original medicine box substrate to which the tag was designed, and the medicine box cardboard, the film layer, and the adhesive layer together determine the electromagnetic field distribution near the tag; furthermore, it calls the free space loss model to describe the basic attenuation law of electromagnetic waves from the reader to the tag and back under the condition of no significant obstruction and multipath disturbance; Based on the above parameters, the ideal state generation unit constructs an ideal reference feature vector by combining node spatiotemporal data; the ideal received signal strength attenuation curve indicates that when the tag is completely attached to the original medicine box and there are no obvious reflectors in the reading path, the echo intensity should change smoothly as the tag enters, passes through and leaves the reading area; the ideal radio frequency phase angle expected value indicates that the propagation path and antenna electrical length between the tag and the reader should remain continuous and stable under ideal conditions. For example, two boxes of vaccines correspond to label identification code A and label identification code B respectively. Although the electronic product codes may belong to the same product batch, due to slight differences in chip batch and packaging details, the system calls parameter set M1 for label identification code A and parameter set M2 for label identification code B, and generates ideal curves B1 and B2 respectively under the same pharmacy node, so that each box of medicine can be compared with its own expected state. Furthermore, as an anomaly handling mechanism, if a certain tag identification code has not yet been mapped in detail in the parameter library, the system can revert to the standard parameter set of the same type of tag for reconstruction and add a low-confidence benchmark mark; if the node site environment undergoes significant changes, the system can switch to a conservative benchmark mode and only output a more lenient ideal reference range; if the base material parameters required for ideal state generation are missing, fine judgment will not be performed, but the system will be transferred to waiting for re-sampling or manual verification. A vaccine box is affixed to the outside of a standard cardboard box, and its backing material has been registered as having a specific dielectric range. When the vaccine box arrives at the hospital, the system retrieves the chip parameters based on the label identification code and reconstructs a smooth ideal received signal strength attenuation curve and a continuous ideal phase expectation by combining the fixed gantry position of the pharmacy receiving node. If the deviation of the on-site sampling is greater than the preset deviation threshold, the system further analyzes the source of the deviation. The purpose of this embodiment is to establish an ideal reference frame with individual targeting for each label.

[0022] Furthermore, the mutation simulation module includes a structural damage mutation unit, a cloning mutation unit, and a simulation output unit. The structural damage mutation unit is used to call a preset antenna micro-deformation impedance mutation operator and a preset backplane dielectric constant step factor, and inject them into the corresponding dimension of the ideal reference feature vector to change the expected value of the ideal RF phase angle and the ideal received signal strength indication attenuation curve. The cloning mutation unit is used to call a preset non-original chip capacitance tolerance factor and inject it into the ideal reference feature vector to change the tag chip resonance characteristics. The simulation output unit is used to generate a damaged simulation feature vector based on the injection results of the structural damage mutation unit and the cloning mutation unit.

[0023] This embodiment provides a refinement mechanism for a mutation simulation module; an ideal benchmark alone is insufficient to identify complex anomalies, because there are many reasons for deviations from the ideal state in the field. If the physical characteristics of the tampering behavior are lacking, the system may still misjudge all deviations as noise; therefore, this embodiment maps the tampering behavior to a calculable and comparable physical mutation trajectory. The structural damage variation unit mainly corresponds to situations where the label is torn off, bent, partially stretched, or reattached; the antenna of the medicine box label is usually printed on a flexible substrate. Once peeled off and reattached, the conductive pattern may develop micro-cracks, geometric changes, or local stress concentrations, altering the impedance matching state; the antenna micro-deformation impedance mutation operator is used to introduce the phenomenon of originally continuous matching but now exhibiting local mismatch into an ideal reference; the backplate dielectric constant step factor is used to reflect the change in electromagnetic boundary conditions caused by the abrupt change in the adjacent dielectric environment after the label is transferred from the original cardboard box to other packaging materials; when both work together, they will simultaneously cause a shift in the expected phase value and a change in the shape of the received signal strength indication attenuation curve; The clone mutation unit mainly corresponds to the situation of illegally copied tags; counterfeiters may copy the surface code, but it is difficult to make the parasitic capacitance, input matching and resonance behavior of the substitute chip completely equivalent to the original device; the non-original chip capacitance tolerance factor is used to express the matching of digital surface identification data, but the chip electrical response deviation is within the preset tolerance range. Its change does not necessarily show a significant decrease in intensity, and is sometimes more easily observed in phase trajectory, resonance stability or scattering consistency. The simulation output unit outputs the above two types of variation results as damaged simulation feature vectors. For example, in the same simulation process, based on the ideal baseline feature vector B at the starting point, the first damaged simulation feature vector D1 is obtained after structural damage injection, and the second damaged simulation feature vector D2 is obtained after clone injection. The system can compare the field residuals with the theoretical residuals corresponding to D1 and D2 respectively, or it can superimpose two or more types of abnormal factors on the same ideal baseline B to form a combined damaged simulation feature vector D3, which is used to identify composite anomalies. Furthermore, as an anomaly handling mechanism, if the on-site operations are known to involve legitimate box replacement, subsidy labeling, or rework and relabeling processes, the system will call the process whitelist to avoid misjudging controlled process operations as malicious sabotage; if a new type of forgery method has not yet been modeled in the simulation library, the system will retain the unknown offset state and transfer it to pending resampling or manual identification; if both the simulation results for structural damage and cloning are partially similar to the on-site residuals but the evidence is insufficient, a conservative result of suspected anomaly and pending supplementary sampling can be output. During the acceptance process at the hospital pharmacy, the numerical content on the label of a box of vaccines was consistent with the system record, but a local abrupt change occurred in the phase trajectory at the scene. After the system called the antenna micro-deformation simulation and the non-original chip tolerance simulation respectively, it was found that the damage characteristics generated by the former were closer to the waveform at the scene, indicating that the label was more likely to have been torn off and re-attached. The purpose of this embodiment is to transform the abstract risk of tampering into a quantifiable and comparable physical anomaly model.

[0024] Furthermore, the residual generation module includes a measured residual extraction unit, a simulation residual extraction unit, and a synchronization normalization unit. The measured residual extraction unit is used to perform differential processing on the field multi-dimensional acquisition vector and the ideal benchmark feature vector to generate a measured residual feature set. The simulation residual extraction unit is used to perform differential processing on the damaged simulation feature vector and the ideal benchmark feature vector to generate a simulation residual feature set. The synchronization normalization unit is used to perform time alignment and scale normalization processing on the measured residual feature set and the simulation residual feature set to generate a comparable residual feature set.

[0025] This embodiment provides a refinement mechanism for the residual generation module. If the original waveform on site is directly compared with the simulated waveform, it is still easily affected by differences in reading distance, dwell time and sampling rhythm. Especially when the vaccine transport box passes through the gantry antenna, the different moving speeds of different boxes may cause changes in the length and amplitude of the reading curve. Therefore, this embodiment extracts residuals through both measured residuals and simulated residuals, and eliminates irrelevant scale differences through synchronous normalization. The measured residual extraction unit performs differential processing on the field multidimensional acquisition vector and the ideal reference feature vector. Its physical meaning is to subtract the ideal part that should exist in the field observation and retain all components that deviate from the ideal state. For example, if the tag should show a smooth intensity change of strengthening and then weakening when passing through the gantry in the ideal state, but the field shows local signal attenuation in the middle section, then this local attenuation feature will be highlighted in the measured residual. If a local inflection point occurs in the originally continuous phase change, the inflection point will also be retained in the residual. The simulation residual extraction unit differs the damaged simulation feature vector with the ideal baseline feature vector to obtain the theoretical abnormal offset profile; this simulation residual is equivalent to an abnormal feature template, focusing on how much and what form the deviation will take after a certain type of tampering occurs. The synchronization normalization unit is used to perform time alignment and scale normalization on two types of residuals. Time alignment solves the problem of differences in the sampling sequence length of the same anomaly under different sampling rhythms. Scale normalization weakens the overall amplitude differences caused by distance, angle, and power settings, making subsequent comparisons focus more on morphological structure rather than absolute values. For example, the measured residual segment R1 extracted in the same reading session may contain 5 sampling points, while the simulated residual segment R2 corresponding to the anomaly template selected in that session may contain 8 sampling points, and there is a difference in the original sampling length between the two. After time alignment, the measured residual segment R1 and the simulated residual segment R2 are mapped to a comparison interval of the same length. If the amplitude of the R1 signal is too high only because the reader reading distance is shortened, the peak and valley positions and phase transition trends can still be preserved after normalization. Furthermore, to avoid the residual generation process being too abstract, this embodiment provides a structured definition for the processing rules: first, the received signal strength indication residual, phase residual, and scattering residual are extracted according to the feature dimensions, and then a multidimensional residual sequence is formed according to the timestamp order within the same reading session; if a certain dimension has missing measurements, the missing measurement mark of that dimension is retained, and other dimension values ​​are not forcibly replaced. During time alignment, the measured residuals and simulated residuals are mapped to a unified comparison interval, with the start and end times of the reading session as the boundary. If there are short-term breaks in the field segments, only continuous valid segments are locally aligned, and low-confidence markers are added to the broken segments. During scale normalization, overall bias and unified amplitude dimensions are eliminated first, so that the residuals at different nodes and different reading distances fall into the comparable interval, but structural information such as peak positions, inflection points, and phase reversal directions used to identify tampering mechanisms are not smoothed out. In other words, the normalization process compresses irrelevant overall magnitude differences and retains morphological differences related to the source of physical anomalies. Furthermore, when multiple anomaly simulation templates exist simultaneously, the synchronous normalization unit can generate a set of comparable residual feature sets for each simulation template and send the results of each set in parallel to the subsequent coupled decision module, instead of retaining only a single template in advance. This can avoid discarding candidate anomaly types too early in the residual stage and is also conducive to comparing the degree of fit between structural damage, clones and composite anomalies in the subsequent stage. Furthermore, to maintain strict consistency with the comparable residual feature vectors in the subsequent coupling decision module, this embodiment provides a unified explanation of the rules for generating comparison inputs from comparable residual feature sets: each set of comparable residual feature sets output by the synchronization normalization unit corresponds to the same reading session and the same anomaly template; before being sent to the coupling decision module, the system unfolds and concatenates the received signal strength indication residual sequence, phase residual sequence, and scattering residual sequence in the feature set into a comparable residual feature vector in a fixed order. If there is a missing dimension, a missing marker is written at the corresponding position and the weight attenuation information is recorded synchronously; thus, the residual generation module outputs a set-organized result, and the coupling decision module uses a vectorized input derived from this set-organized result. The two have the same source, the same boundary, and the same session; Furthermore, the above vectorization process does not change the physical semantics of the residuals, but only converts the original multidimensional residual sequence organized by dimension and time sequence into a unified input format that is easy for the algorithm to call. For scenarios using dynamic time warping algorithms, the comparable residual feature vectors can be understood as multidimensional time sequence inputs encoded according to unified rules. For scenarios using cosine similarity algorithms, they can be understood as comparison vectors of fixed length. Regardless of the algorithm used, the same set of comparable residual features is used as the data source, thereby avoiding inconsistencies in the naming and boundaries of the comparison objects in the preceding and following steps. In abnormal handling situations, if the on-site sampling is too short to extract a residual structure sufficient to reflect the trend, the system will not perform high-confidence similarity analysis and will directly proceed to the re-sampling stage. If there are obvious breakpoints in the on-site data, such as a forklift passing by causing instantaneous shielding, the synchronization normalization unit can remove the abnormal segment or reduce its weight. If there are multiple candidate templates in the simulation residual library, the system can simultaneously retain multiple sets of comparable residual feature sets for optimal matching in the subsequent coupling decision stage. Furthermore, if the number of effective sampling points in the same session reaches the minimum but only covers a single-sided segment of the gantry process, such as only sampling the entry beam stage and not the exit beam stage, the system can continue to generate residuals, but will add a single-sided sampling mark to the result, so that the subsequent decision will prioritize the use of conservative thresholds to avoid amplifying the probability of misjudgment due to incomplete segments. When receiving a box of vaccines at a hospital pharmacy, the box was quickly pushed into the gantry. One label only formed a short on-site reading fragment, while the abnormal residual fragment of tearing and re-adhesion obtained by the system simulation was longer. After synchronization and normalization, the two were mapped to the same comparison scale. The results showed that the phase residuals showed similar mutation profiles in the middle and later stages. Although the other label had a lower intensity, after normalization, it only showed an overall sinking and did not have any local structure consistent with any tampered template. The purpose of this embodiment is to transform the original observational differences into a directly comparable representation of anomalies.

[0026] Furthermore, the coupling decision module includes a similarity calculation unit, an environmental coupling degree calculation unit, and a state determination unit; the similarity calculation unit is used to calculate the comparable residual feature set based on the multidimensional dynamic time warping algorithm or the cosine similarity algorithm to generate a tampering coupling score. The environmental coupling calculation unit is used to calculate the measured residual feature set based on the preset multipath fading feature template and the preset occlusion attenuation feature template to generate an environmental coupling score. The state determination unit is used to compare the tampering coupling score with a preset tampering threshold and the environmental coupling score with a preset environmental threshold. When the tampering coupling score is greater than or equal to the tampering threshold, the physical tampering decision result is output; when the tampering coupling score is less than the tampering threshold and the environmental coupling score is greater than or equal to the environmental threshold, the environmental noise decision result is output. When the tampering coupling score is less than the tampering threshold and the environmental coupling score is less than the environmental threshold, the decision result for re-collection is output.

[0027] This embodiment provides a refined mechanism for the coupled decision module; relying solely on whether there is a shift in the residual is insufficient to draw a reliable conclusion, as the shift may originate from tag anomalies or environmental complexity; if tampering similarity and environmental similarity are not modeled separately, multipath disturbances caused by metal reflection may be misjudged as physical tampering, or hidden physical damage to tags may be concealed in environmental noise; therefore, this embodiment decomposes the decision into three parts: tampering coupling analysis, environmental coupling analysis, and state determination. The similarity calculation unit is used to measure the degree of morphological fit between comparable residual feature sets and abnormal simulation residuals. When using multidimensional dynamic time warping, the focus is on comparing whether the change trajectory of multidimensional features remains consistent after appropriate stretching. When using cosine similarity, the focus is on comparing whether the directions of multidimensional residuals are consistent. Its core is not to compare absolute size, but to judge whether the contour structure of the deviation in the field is consistent with the preset tampering mechanism. The environmental coupling degree calculation unit determines whether the measured residuals on site conform to environmental templates such as multipath fading and obstruction attenuation. Multipath fading is usually manifested as periodic fluctuations or local peak-valley alternations caused by the superposition of reflection paths. Obstruction attenuation is usually manifested as an overall decrease in intensity, followed by recovery after passing through, and the phase change reflects the obstruction of the propagation path more than a sudden change in the antenna's electrical characteristics. The status determination unit classifies data based on two types of scores. If the tampering coupling score, which represents the consistency of tampering features, reaches the tampering threshold, it indicates that the on-site offset is highly consistent with a certain type of tampering feature template. In this case, even if the environmental coupling score, which represents the consistency of environmental disturbances, is high, the physical tampering decision is still output first. If the tampering coupling score does not reach the threshold but the environmental coupling score reaches the environmental threshold, it indicates that the offset can be mainly explained by the environment. The system maintains the business chain without interruption and corrects the reading conditions. If both the tampering coupling score and the environmental coupling score are insufficient, the system enters the waiting-for-re-collection state. Furthermore, to make the judgment boundaries clearer, this embodiment limits the use of the scoring method: the tampering coupling score is used to answer whether the on-site residuals are sufficient to be explained by a certain type of anomaly mechanism, and the environmental coupling score is used to answer whether the on-site residuals are sufficient to be explained by the propagation of environmental disturbances. The two types of scores are modeled and thresholded separately, and are not simply subtracted to cancel each other out. Specifically, a higher environmental coupling score does not directly cancel out the tampering coupling score. Instead, it is first determined whether the tampering evidence has reached a level that can be independently established, and then it is determined whether the remaining situation is more suitable to be attributed to the environment. This can avoid the concealment of tampering being mistakenly covered by environmental interference characteristics when it happens to occur in a complex environment. Furthermore, when the same site residual is similar to multiple tampered templates, the similarity calculation unit can output multiple candidate tampering coupling scores, and use the highest score as the tampering coupling score for this session, while recording the corresponding anomaly type label; if the highest score and the second highest score are close and both are in a high position, the system can retain a suspected composite anomaly mark while outputting the physical tampering conclusion, so as to trace the more specific source of the anomaly during subsequent manual review; Furthermore, to maintain consistency with the aforementioned residual generation rules, although the similarity calculation unit in this embodiment is described in the embodiment text as calculating the comparable residual feature set, its actual calculation input is the comparable residual feature vector generated from the comparable residual feature set in a unified order; the calculation of the measured residual feature set by the environment coupling degree calculation unit also follows the same dimensional order and time order organization rules; in other words, whether it is tampered template comparison or environment template comparison, it is based on the residual representation formed within the same session and under the same sorting rules, thereby ensuring that the feature set and feature vector are only different names at the data organization level and algorithm input level, rather than comparison objects from different sources; Furthermore, before entering the environmental coupling degree calculation unit, the multipath fading feature template and the occlusion attenuation feature template are also templated and encoded according to the same dimensional arrangement rule as the comparable residual feature vector, so that the environmental template and the tampered template remain isomorphic at the input interface. In this way, on the one hand, it is convenient to compare the interpretability of different types of templates for field residuals under the same scoring framework, and on the other hand, it avoids additional scoring bias due to different template organization methods. In abnormal or boundary situations, if a sudden change in the node environment is detected during the on-site reading process, the environmental coupling score can be analyzed using a short-time template first; if both the tampering coupling score and the environmental coupling score are close to their respective threshold edges, the system can output a conservative strategy of prioritizing re-collection; if multiple consecutive readings of the same tag show moderate tampering coupling but the fluctuation of the environmental score exceeds the preset tolerance, multiple session evidences can be accumulated before making a judgment. Furthermore, the so-called "approaching the threshold edge" can be specifically set in the engineering deployment as falling into a preset buffer zone near the threshold. When entering this zone, the system does not immediately make a final release or a final block, but instead prioritizes triggering resampling, close-range verification, or calling the historical session of the same node for cross-validation, thereby reducing the impact of accidental fluctuations near the threshold boundary on the conclusion. During the outbound process from the regional cold storage, the measured residual of a vaccine label was highly consistent with the simulated template of tearing and re-sticking at the location of the phase change, and this was repeated in the secondary sampling, indicating that the system output the result of physical tampering; another vaccine received a signal strength indicator near a metal turnover cage showing fluctuations exceeding the preset tolerance, but the phase change was more consistent with the multipath fading template, indicating that the system output the result of environmental noise; and yet another vaccine entered the waiting-for-re-sampling state due to the short sampling segment caused by a short-term hand obstruction. The purpose of this embodiment is to distinguish between tampering with similar mechanisms and explainable environmental disturbances through a dual-scoring decision framework.

[0028] Furthermore, the traceability feedback module includes a link control unit, a compensation control unit, and a re-collection control unit; the link control unit is used to generate a traceability chain blocking instruction to prevent the traceability record corresponding to the electronic tag from being written to subsequent traceability nodes in response to the physical tampering judgment result, and to write the electronic tag digital identification data, the underlying radio frequency feature data, and the spatiotemporal data of the traceability node into the abnormal traceability record; The compensation control unit is used to generate, in response to the environmental noise decision result, an RF compensation command based on the environmental coupling score for adjusting at least one of the RFID reader's transmit power, operating frequency, receive gain, or sampling time window, and sends the RF compensation command to the RFID reader; The re-collection control unit is used to generate a re-collection command in response to the re-collection decision result and mark the current status of the traceability node as pending confirmation.

[0029] This embodiment provides a refined mechanism for the traceability feedback module. If the system only provides a result after making a judgment without linking the traceability chain control and adjusting on-site read / write parameters, the judgment cannot truly be transformed into a business closed loop. For example, allowing continued storage after discovering tampering will undermine the credibility of vaccine traceability; identifying environmental noise but not correcting the read / write conditions will lead to the same problem recurring; failing to freeze the node status when there is insufficient evidence may result in defective traceability records. Therefore, this embodiment sets up link control, compensation control, and re-collection control for the three judgment results respectively. Upon receiving a physical tampering judgment, the link control unit immediately generates a traceability chain blocking instruction, preventing the corresponding tag from being written to subsequent nodes. The blocking can manifest as prohibiting the system from recording the medicine box as inspected and received, or as prohibiting it from being scanned and released in subsequent distribution stages. At the same time, the link control unit writes the digital identification data, underlying radio frequency characteristic data, and node spatiotemporal data involved in this anomaly into the anomaly traceability record, preserving an evidence chain for subsequent quality investigations, batch recalls, or law enforcement evidence collection. After receiving the environmental noise decision, the compensation control unit does not directly regard the tag as an anomaly, but generates radio frequency compensation instructions based on the environmental coupling score and noise type. If it is more consistent with the enhanced reflection caused by the proximity of metal or standing wave interference, it can be improved by switching the operating frequency or adjusting the antenna transmit power. If it is more consistent with the instantaneous attenuation caused by obstruction, the sampling window can be appropriately extended or the receiving gain can be increased to obtain a more complete echo segment. After receiving the decision to re-collect data, the re-collection control unit generates a re-collection instruction and marks the current traceability node as pending confirmation. The business system can clearly distinguish between three states: confirmed legal, confirmed abnormal, and pending confirmation, avoiding intermediate states from being mistakenly identified as pass or block. In the pending confirmation state, the operator can be asked to adjust the medicine box posture, shorten the reading and writing distance, change the handheld device, or move to a low-reflection area before re-collecting data. Furthermore, as an anomaly handling mechanism, if a network interruption causes the blocking command to temporarily fail to be synchronized to the upper-level traceability platform, the local controller should first freeze the subsequent operation permissions of the tag at the edge side, and then retransmit the anomaly record after the link is restored; if high environmental coupling continues to occur after the compensation control is executed once, the system can limit the number of compensations, and if the number is exceeded, it will automatically switch to manual review; if the re-sampling exceeds the preset number and still cannot form a stable conclusion, it can be required to unpack and verify at close range or replace the reading and writing device; At the hospital pharmacy acceptance site, a box of vaccines showed a phase abrupt change consistent with the tearing and relabeling after two readings. The system immediately blocked the entry of the box of medicine into the inventory and saved its digital code, abnormal radio frequency segment, and acceptance station location information to the abnormal record. Another box of vaccines showed significant attenuation only because it was close to the stainless steel receiving truck. After the system judged it to be environmental noise, it automatically adjusted the working frequency and extended the sampling time window. After the second reading returned to normal, it continued to be written into the traceability chain. Another box of vaccines had unstable sampling due to frequent operator movement. The system marked it as pending confirmation and prompted for re-sampling. The purpose of this embodiment is to transform the judgment results into executable traceability control actions.

[0030] The RFID electronic tag data traceability method includes the following steps: S1. Acquire the electronic tag digital identification data, underlying radio frequency feature data, and traceability node spatiotemporal data collected by the RFID reader to generate a multi-dimensional acquisition vector on site. S2. Call the preset ideal radio frequency digital twin model, and based on the tag identification code and traceability node spatiotemporal data in the field multidimensional acquisition vector, perform benchmark reconstruction on the field multidimensional acquisition vector to generate an ideal benchmark feature vector. S3. Based on the preset abnormal state simulation model, physical mutation injection is performed on the ideal baseline feature vector to generate the damaged simulation feature vector. S4. Based on the on-site multi-dimensional acquisition vector, ideal benchmark feature vector and damaged simulation feature vector, generate the measured residual feature set and the simulation residual feature set, and perform time alignment and scale normalization processing to generate a comparable residual feature set. S5. Calculate the comparable residual feature set, calculate the measured residual feature set based on the preset feature template to generate the environmental coupling score, and combine the environmental coupling score to generate the traceability validity judgment result. S6. Based on the traceability validity judgment result, generate traceability chain blocking instruction, radio frequency compensation instruction, or re-acquisition instruction.

[0031] This embodiment provides a step mechanism for an RFID electronic tag data traceability method, such as... Figure 2 As shown; this method can be executed by servers or edge controllers deployed in vaccine manufacturing enterprises, regional cold storage facilities, transportation handover points, and hospital pharmacies. Its process corresponds one-to-one with the aforementioned system structure, but it places greater emphasis on completing traceability and identification in chronological order. In step S1, the system collects the tag's digital identifier, underlying radio frequency characteristics, and node spatiotemporal information at the current node to form a multi-dimensional acquisition vector on site. For the cold chain vaccine scenario, this step solves the problem of completely recording the identity and physical state of a tag at a specific node and at a specific time. In step S2, the system calls the ideal radio frequency digital twin model based on the tag identification code and node spatiotemporal data to obtain the ideal baseline feature vector. This step is not simply finding a historical average curve, but rather combining the tag individual identity and node scenario to reconstruct its expected response when it is undamaged, unreposted, and uncloned. In step S3, the system injects physical variations into the ideal baseline based on the abnormal state simulation model to form a damaged simulation feature vector; this can transform potential risks such as labels being torn off and reattached and chips being replaced or cloned into specific and comparable physical anomaly templates. In step S4, the system constructs the measured residual and the simulated residual respectively, and performs time alignment and scale normalization on the two; its essence is to transform the original data into an abnormal expression that deviates from the ideal state, while eliminating the comparison obstacles caused by the difference in sampling rhythm and overall amplitude. In step S5, the system calculates the similarity of comparable residual feature sets and generates a traceability validity judgment result by combining the environmental coupling score; the conclusion may be one of three categories: physical tampering, environmental noise, or need for re-sampling. In step S6, the system outputs different instructions based on the judgment result; if physical tampering is confirmed, the tracing chain is blocked; if it is identified as environmental noise, radio frequency compensation is issued; if the evidence is insufficient, re-collection is required. For example, when a box of vaccines generates a multi-dimensional acquisition vector at a hospital node, the system generates an ideal baseline feature vector via S2, obtains a damaged simulation feature vector via S3, and forms a measured residual feature set and a simulation residual feature set via S4. If S5 determines that the measured residual is closer to the simulation residual, then S6 outputs a blocking command. If the measured residual is closer to the environmental template, then a compensation command is output. If neither is sufficient, then a re-acquisition command is output. Furthermore, as an anomaly handling mechanism, if the data is incomplete in stage S1, the process can be paused at the acquisition end and will not directly proceed to subsequent analysis; if the model parameters are missing in stage S2 or S3, the system can switch to conservative mode and only output prompts for re-collection or manual verification; if the two types of scores hover in the boundary range for a long time in stage S5, the results of the most recent sampling sessions can be superimposed for a comprehensive judgment. During the process from the regional cold storage outbound to the transportation handover, when a box of vaccines passes through the gantry, the system executes steps S1 to S6. For most of the boxes, the on-site offset is identified as having strong environmental coupling in step S5, so the system automatically adjusts the reading parameters in step S6 and successfully completes the outbound registration. However, for a few boxes, the on-site residual and the abnormal template of transfer damage maintain a high degree of consistency in multiple sampling periods, so the system blocks the box from continuing to enter the transportation handover stage in step S6. The purpose of this embodiment is to achieve closed-loop control of the entire chain of traceability data from collection, reconstruction, simulation, comparison to disposal in a process-oriented manner. Furthermore, S3 includes: calling a preset antenna micro-deformation impedance abrupt change operator and a preset backplane dielectric constant step factor, and injecting the antenna micro-deformation impedance abrupt change operator and the backplane dielectric constant step factor into an ideal reference feature vector to generate structural damage simulation features; calling a preset non-original chip capacitance tolerance factor and injecting the non-original chip capacitance tolerance factor into the ideal reference feature vector to generate clone damage simulation features; and generating a damage simulation feature vector based on the structural damage simulation features and the clone damage simulation features.

[0032] This embodiment provides a refined mechanism for step S3. If S3 only generates an abnormal simulation state in a general way, although it can express the existence of abnormal simulation, it cannot explain the physical differences of the abnormal source, and it is not conducive to review and reproduction in actual deployment. In order to make the simulation process correspond one-to-one with the real tampering mechanism, this embodiment further divides S3 into two paths: structural damage simulation and clone damage simulation. The structural damage simulation targets scenarios where the tag experiences mechanical disturbances and abrupt changes in attachment conditions. The antenna micro-deformation impedance mutation operator reflects the mismatch effect caused by tearing, bending, or local cracks in the conductive pattern. The backplane dielectric constant step factor reflects the impact of sudden changes in the surrounding electric field boundary conditions after the tag is transferred from the original medicine box substrate to another packaging surface. After injecting both into an ideal reference, the structural damage simulation characteristics can be obtained. The common manifestations are non-smooth transitions in the phase trajectory and the intensity attenuation law no longer maintaining the continuity under the ideal state. Among them, the antenna micro-deformation impedance mutation operator is achieved by matrix addition of a preset phase shift bias matrix and the expected value of the ideal radio frequency phase angle; the backplane dielectric constant step factor is achieved by multiplying the slope of the ideal received signal strength indication attenuation curve by a preset attenuation compensation coefficient. Specifically, the phase shift offset matrix is ​​a phase offset constant matrix extracted by least squares fitting based on the measured phase data of the RFID electronic tag under known peeling and reattaching conditions; the attenuation compensation coefficient is an empirical value pre-calibrated based on the ratio of the relative permittivity of the target packaging material to the original packaging material. Cloned damage simulation is designed for scenarios involving illegal chip replacement or copying of tag codes; the non-original chip capacitance tolerance factor is used to express the inherent differences in input matching, parasitic parameters, and resonance characteristics of the counterfeit chip; the cloned damage simulation features it forms are often not large-scale mechanical distortions, but rather subtle drifts at the electrical characteristic level, making it particularly suitable for identifying anomalies where the digital surface is legitimate but the physical underlying structure is inconsistent. Based on this, the system generates a damage simulation feature vector based on structural damage simulation features and clone damage simulation features. For the same ideal baseline feature vector, structural damage simulation features and clone damage simulation features can be generated separately. If a single anomaly is suspected on-site, the on-site residuals can be compared with the theoretical residuals corresponding to these two types of simulation features. If the business risk is high, the structural damage simulation features and clone damage simulation features can also be combined to form a composite anomaly damage simulation feature, which can be used to identify the situation of removing the label and replacing the chip. From an evolutionary perspective, without this refined breakdown, all anomalies might be compressed into the same fuzzy template, resulting in insufficient ability to distinguish between different anomalies. Therefore, this embodiment explicitly introduces two types of physical factors to provide a clearer basis for identifying the physical sources of different anomalies in the future. Furthermore, as an anomaly handling mechanism, if there is a legitimate rework and re-attachment process for the medicine box on site, the structural damage simulation should be constrained by the rework process record; if the newly discovered cloned chip model exceeds the existing factor library, the system can first generate an approximate simulation using the parameter range of adjacent devices and mark the approximate template; if neither type of simulation can explain the on-site anomaly, the unknown anomaly type is retained and the manual sample reinjection process is initiated. During the acceptance process at the hospital pharmacy, the electronic product code, label identification code, and verification code of a certain box of vaccines were consistent with the records, but the residual at the site showed a stable inflection point in the local phase. The system first generated corresponding simulation features according to the structural damage path, and then generated corresponding simulation features according to the cloning path. After comparison, it was found that the anomaly at the site was closer to the simulation features of structural damage, and the physical variation features of the corresponding label were consistent with the structural damage model of peeling off and re-attaching the original packaging. Another box of vaccines did not show obvious mechanical waveform distortion, but the resonance performance was inconsistent with the original chip. The system was more inclined to match the simulation features of cloning damage. The purpose of this embodiment is to refine the anomaly simulation in step S3 into a dual-path mechanism that can distinguish the source of the anomaly.

[0033] Furthermore, S5 and S6 include: calculating comparable residual feature sets based on a multidimensional dynamic time warping algorithm or a cosine similarity algorithm to generate a tampering coupling score; calculating measured residual feature sets based on a preset multipath fading feature template and a preset occlusion attenuation feature template to generate an environmental coupling score; generating a physical tampering decision result and a traceability chain blocking instruction when the tampering coupling score is greater than or equal to a preset tampering threshold; generating an environmental noise decision result and an RF compensation instruction when the tampering coupling score is less than the tampering threshold and the environmental coupling score is greater than or equal to an environmental threshold; and generating a re-acquisition decision result and a re-acquisition instruction when the tampering coupling score is less than the tampering threshold and the environmental coupling score is less than the environmental threshold.

[0034] This embodiment provides a joint refinement mechanism for steps S5 and S6. If the judgment in S5 is separated from the handling in S6, the system may encounter inconsistencies such as identifying environmental noise in the analysis but still blocking the flow of business, or there may be a delay risk of suspected tampering but not immediately freezing the traceability chain. Therefore, this embodiment describes the judgment and execution in a linked manner, so that each type of scoring result corresponds to a clear action output. The system calculates comparable residual feature sets based on a multidimensional dynamic time warping algorithm or a cosine similarity algorithm to generate a tampering coupling score. The comparison focuses on the consistency of the field residuals with the theoretical anomaly template in terms of multidimensional morphology. The system calculates measured residual feature sets based on multipath fading templates and occlusion attenuation templates to generate an environmental coupling score. The two types of scores respectively evaluate whether the field residuals meet the tampering characteristics and whether they meet the environmental interference characteristics. The system directly triggers the handling result according to the preset decision logic; if the tampering coupling score reaches or exceeds the tampering threshold, a physical tampering judgment result is generated and a traceability chain blocking instruction is output at the same time; if the tampering coupling score does not reach the threshold but the environmental coupling score reaches or exceeds the environmental threshold, an environmental noise judgment result is generated and a radio frequency compensation instruction is output at the same time; if both types of scores are insufficient, a re-collection judgment result is generated and a re-collection instruction is output. For example, in a certain read session, if the system output tampering coupling score reaches the tampering threshold but the environmental coupling score does not reach the environmental threshold, the system executes the physical tampering handling path, directly freezing the subsequent writes of the tag after the current pharmacy node; if the system output tampering coupling score does not reach the tampering threshold but the environmental coupling score reaches the environmental threshold, the system executes the environmental noise handling path, sending a parameter update packet to the reader and retrying; if neither score reaches the corresponding threshold, the system executes the pending re-collection handling path, only generating a re-collection task without making a final conclusion of allowing or blocking; through this method of one-time judgment and direct mapping to a class of actions, the latency and ambiguity caused by manual intermediate judgment can be avoided; From the perspective of evolutionary reinforcement, setting only the tampering threshold will cause the system to be biased towards conservative interception in complex cold chain environments, affecting normal business operations; setting only the environmental threshold may allow covert tampering to infiltrate under the cover of environmental fluctuations; therefore, this embodiment sets two types of thresholds and three types of output paths at the same time, so that the system can protect traceability security while also taking into account on-site availability. Furthermore, as an anomaly handling mechanism, if the score of a session is in the threshold border zone, the system can require two or three consecutive sampling results to remain on the same branch before executing the final action; if the on-site environment does not improve after the compensation instruction is issued, compensation will not be unlimited, but will be transferred to waiting for resampling or manual review after reaching a preset number of times; if the blocking instruction has been issued but subsequent manual verification proves that it is a legitimate rework or node configuration error, the system can perform unlocking and audit follow-up based on the anomaly record; During peak acceptance at the hospital pharmacy, the residual data of one vaccine box showed consistent consistency with the transfer damage template in both phase and scattering dimensions. The system directly generated a physical tampering conclusion and blocked its entry into the database. The residual data of another vaccine box mainly showed overall attenuation and periodic fluctuations, which was more consistent with the multipath fading template. The system issued compensation instructions for frequency modulation and extended sampling window and then read the data again. Another vaccine box only had scattered residual fragments due to the operator moving too quickly. The system could not confirm whether it was tampered with or affected by environmental interference, so it marked the box as pending confirmation and prompted for re-collection. The purpose of this embodiment is to directly couple the judgment logic and the handling logic to achieve a unified system of rapid response, cautious release, and abnormal interception in complex situations.

[0035] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An RFID electronic tag data traceability system, characterized in that, include: The data acquisition module is used to acquire the electronic tag digital identification data containing the tag identification code, the underlying radio frequency feature data, and the spatiotemporal data of the traceability node collected by the RFID reader, so as to generate a multi-dimensional acquisition vector on site. The benchmark reconstruction module is used to call a preset ideal radio frequency digital twin model, and based on the tag identification code and traceability node spatiotemporal data in the field multidimensional acquisition vector, to perform benchmark reconstruction on the field multidimensional acquisition vector to generate an ideal benchmark feature vector. The mutation simulation module is used to inject physical mutations into the ideal baseline feature vector based on a preset abnormal state simulation model to generate a damaged simulation feature vector. The residual generation module is used to generate a measured residual feature set and a simulated residual feature set based on the on-site multidimensional acquisition vector, the ideal benchmark feature vector, and the damaged simulation feature vector, and to perform time alignment and scale normalization processing to generate a comparable residual feature set. The coupling decision module is used to calculate the similarity of the comparable residual feature set to generate a tampering coupling score, calculate the measured residual feature set based on a preset feature template to generate an environmental coupling score, and combine the environmental coupling score to generate a physical tampering decision result, an environmental noise decision result, or a decision result to be re-sampled. The traceability feedback module is used to generate a traceability chain blocking instruction in response to the physical tampering judgment result, generate an radio frequency compensation instruction in response to the environmental noise judgment result and send it to the RFID reader, or generate a re-collection instruction in response to the re-collection judgment result.

2. The RFID electronic tag data traceability system according to claim 1, characterized in that, The data acquisition module includes: The digital identification acquisition unit is used to acquire the electronic product code, tag identification code and encrypted anti-counterfeiting verification code from the electronic tag to generate the digital identification data of the electronic tag; The radio frequency feature acquisition unit is used to acquire the received signal strength indication time series, radio frequency phase angle and antenna backscattering cross section parameters from the RFID reader to generate the underlying radio frequency feature data; The node spatiotemporal acquisition unit is used to acquire the spatial coordinates and timestamps of the traceability node in order to generate the spatiotemporal data of the traceability node.

3. The RFID electronic tag data traceability system according to claim 1, characterized in that, The benchmark reconstruction module includes: The model parameter calling unit is used to call the tag identification code, the preset ideal dielectric constant of the tag backing plate, and the preset free space loss model in the electronic tag digital identification data. An ideal state generation unit is used to generate the ideal reference feature vector based on the tag identification code, the spatiotemporal data of the traceability node, the ideal dielectric constant of the tag backplane, and the free space loss model. The ideal reference feature vector includes the ideal received signal strength attenuation curve and the expected value of the ideal radio frequency phase angle.

4. The RFID electronic tag data traceability system according to claim 3, characterized in that, The mutation simulation module includes: The structural damage variation unit is used to call the preset antenna micro-deformation impedance mutation operator and the preset backplane dielectric constant step factor, and inject them into the corresponding dimension of the ideal reference feature vector to change the expected value of the ideal radio frequency phase angle and the ideal received signal strength indication attenuation curve. The clone mutation unit is used to call the preset non-original chip capacitance tolerance factor and inject it into the ideal reference feature vector to change the resonant characteristics of the tag chip; The simulation output unit is used to generate the damaged simulation feature vector based on the injection results of the structural damage mutation unit and the clonal mutation unit.

5. The RFID electronic tag data traceability system according to claim 1, characterized in that, The residual generation module includes: The measured residual extraction unit is used to perform differential processing on the field multidimensional acquisition vector and the ideal benchmark feature vector to generate the measured residual feature set; The simulation residual extraction unit is used to perform differential processing on the damaged simulation feature vector and the ideal reference feature vector to generate the simulation residual feature set; The synchronization and normalization unit is used to perform time alignment and scale normalization on the measured residual feature set and the simulated residual feature set to generate the comparable residual feature set.

6. The RFID electronic tag data traceability system according to claim 5, characterized in that, The coupling decision module includes: The similarity calculation unit is used to calculate the comparable residual feature set based on the multidimensional dynamic time warping algorithm or the cosine similarity algorithm to generate the tampering coupling score. An environmental coupling calculation unit is used to calculate the measured residual feature set based on a preset multipath fading feature template and a preset occlusion attenuation feature template to generate an environmental coupling score. A state determination unit is used to compare the tampering coupling score with a preset tampering threshold and the environmental coupling score with a preset environmental threshold. Specifically, when the tampering coupling score is greater than or equal to the tampering threshold, a physical tampering judgment result is output; When the tampering coupling score is less than the tampering threshold and the environment coupling score is greater than or equal to the environment threshold, the environment noise decision result is output. When the tampering coupling score is less than the tampering threshold and the environmental coupling score is less than the environmental threshold, the decision result for re-sampling is output.

7. The RFID electronic tag data traceability system according to claim 6, characterized in that, The traceability feedback module includes: The link control unit is used to generate a traceability chain blocking instruction to prevent the traceability record corresponding to the electronic tag from being written to subsequent traceability nodes in response to the physical tampering judgment result, and to write the electronic tag digital identification data, the underlying radio frequency feature data and the traceability node spatiotemporal data into the abnormal traceability record. A compensation control unit is configured to, in response to the environmental noise decision result, generate an RF compensation command based on the environmental coupling score for adjusting at least one of the RFID reader's transmit power, operating frequency, receive gain, or sampling window, and send the RF compensation command to the RFID reader; The re-collection control unit is used to generate a re-collection command in response to the re-collection decision result and mark the current status of the traceability node as pending confirmation.

8. A method for tracing data from RFID electronic tags, characterized in that, Includes the following steps: S1. Acquire the electronic tag digital identification data, underlying radio frequency feature data, and traceability node spatiotemporal data collected by the RFID reader to generate a multi-dimensional acquisition vector on site. S2. Call the preset ideal radio frequency digital twin model, and based on the tag identification code and traceability node spatiotemporal data in the field multidimensional acquisition vector, perform benchmark reconstruction on the field multidimensional acquisition vector to generate an ideal benchmark feature vector. S3. Based on the preset abnormal state simulation model, physical mutation injection is performed on the ideal benchmark feature vector to generate a damaged simulation feature vector. S4. Based on the on-site multidimensional acquisition vector, the ideal benchmark feature vector, and the damaged simulation feature vector, generate the measured residual feature set and the simulation residual feature set, and perform time alignment and scale normalization processing to generate a comparable residual feature set. S5. Calculate the comparable residual feature set, calculate the measured residual feature set based on the preset feature template to generate an environmental coupling score, and combine the environmental coupling score to generate a traceability validity judgment result. S6. Based on the traceability validity judgment result, generate a traceability chain blocking instruction, radio frequency compensation instruction, or re-acquisition instruction.

9. The RFID electronic tag data traceability method according to claim 8, characterized in that, S3 includes: The preset antenna micro-deformation impedance mutation operator and the preset backplate dielectric constant step factor are invoked, and the antenna micro-deformation impedance mutation operator and the backplate dielectric constant step factor are injected into the ideal reference feature vector to generate structural damage simulation features. The preset non-original chip capacitor tolerance factor is invoked and injected into the ideal reference feature vector to generate cloned damaged simulation features; Based on the structural damage simulation features and the clone damage simulation features, the damage simulation feature vector is generated.

10. The RFID electronic tag data traceability method according to claim 8, characterized in that, S5 and S6 include: The comparable residual feature set is calculated based on the multidimensional dynamic time warping algorithm or the cosine similarity algorithm to generate the tampered coupling score. Based on the preset multipath fading feature template and the preset occlusion attenuation feature template, the measured residual feature set is calculated to generate an environmental coupling score. When the tampering coupling score is greater than or equal to the preset tampering threshold, a physical tampering judgment result and the traceability chain blocking instruction are generated; When the tampering coupling score is less than the tampering threshold and the environmental coupling score is greater than or equal to the preset environmental threshold, an environmental noise decision result and the radio frequency compensation instruction are generated. When the tampering coupling score is less than the tampering threshold and the environmental coupling score is less than the environmental threshold, a decision result for re-collection and a re-collection instruction are generated.

Images