An online detection and analysis system for traditional Chinese medicine ingredients and system

Abstract

The application discloses a traditional Chinese medicine component online detection and analysis system and a system, and relates to the technical field of traditional Chinese medicine quality analysis.The system comprises a data acquisition module, a fusion identification module, a consensus verification module and a model optimization module.The data acquisition module is used for acquiring multi-modal data of traditional Chinese medicine samples.The fusion identification module is used for performing fusion analysis on the multi-modal data and directly outputting a structured identification conclusion.The consensus verification module is used for realizing cross-terminal data consensus verification by using an intelligent contract.The model optimization module is used for realizing continuous co-evolution of a global identification model.The application significantly improves the comprehensiveness and accuracy of traditional Chinese medicine identification by means of multi-modal data fusion and deep learning analysis, provides tamper-proof, traceable and cross-terminal mutually-recognized credible records for detection results, greatly enhances the public credibility and legal effectiveness of the results, and enables the whole network to have self-adaptive and collective intelligent improvement capabilities, so that the core bottlenecks of information isolation, model solidification, lack of credible cooperation and evolution power in the prior art are systematically solved.

Description

Technical Field

[0001] This invention relates to the field of traditional Chinese medicine quality analysis technology, and in particular to an online detection and analysis system for traditional Chinese medicine components. Background Technology

[0002] The identification of authenticity, evaluation of origin, and assessment of processing techniques for traditional Chinese medicine (TCM) are core aspects of ensuring its clinical efficacy and medication safety. Currently, relevant testing technologies mainly rely on large-scale laboratory analytical instruments (such as high-performance liquid chromatography and mass spectrometry) combined with the experience and judgment of professionals. This approach suffers from problems such as cumbersome procedures, long processing times, high costs, and high professional requirements for operators, making it difficult to meet the needs of rapid screening in the TCM market and online quality control by production enterprises for on-site or near-on-site testing. With the development of sensing technology and artificial intelligence, several technological directions have emerged aimed at achieving rapid detection: On the one hand, portable devices based on near-infrared spectroscopy and Raman spectroscopy are used for the preliminary identification of TCM, but they usually rely only on single physicochemical information, are easily affected by matrix interference, and have limited ability to identify complex characteristics such as origin and processing time; on the other hand, bionic olfaction (electronic nose) and bionic... Taste (electronic tongue) sensors have begun to be applied in the food and pharmaceutical fields, providing the possibility of digitally characterizing the odor and taste features of medicinal materials. However, how to effectively integrate these sensory information with traditional chemical spectra and extract deep-seated, high-value identification features related to the variety, origin, and processing technology of medicinal materials remains a technical challenge that urgently needs to be solved. Existing detection systems are mostly independent "information silos," and their built-in identification models are usually trained based on a limited number of samples. Once deployed, the model performance becomes fixed, making it difficult to continuously evolve and optimize using new and widely distributed data generated in practical applications. At the same time, the detection results of a single terminal lack cross-terminal mutual verification and reliable evidence storage mechanisms. When encountering controversial samples or new counterfeiting methods, the authority and traceability of the results are insufficient, making it difficult to build a reliable and collaborative Chinese medicine quality evaluation network.

[0003] However, current common solutions have many drawbacks, including: existing TCM testing technologies rely on single-dimensional chemical or spectral data, making it difficult to accurately identify complex characteristics such as authenticity and processing techniques; the built-in identification models are static and fixed, unable to continuously evolve with new data to adapt to variations in medicinal materials and new types of adulteration; each testing terminal operates in isolation, lacking a reliable consensus mechanism and secure collaboration path for results, forming "data silos"; there is a fundamental contradiction between data privacy risks and the need for model evolution, and the testing process and results lack tamper-proof end-to-end evidence storage, resulting in weak traceability and legal validity. Summary of the Invention

[0004] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0005] In view of the above-mentioned existing online detection and analysis system for Chinese medicine components and the problems existing in the system, the present invention is proposed.

[0006] Therefore, the purpose of this invention is to provide an online detection and analysis system for traditional Chinese medicine components. This system is applicable to solving the problems of existing traditional Chinese medicine detection technologies that rely on single-dimensional chemical or spectral data, making it difficult to accurately identify complex characteristics such as authenticity and processing techniques; the built-in identification models are static and fixed, unable to continuously evolve with new data to adapt to variations in medicinal materials and new types of adulteration; each detection terminal operates in isolation, lacking a reliable consensus mechanism and secure collaborative path for results, forming "data silos"; there is a fundamental contradiction between data privacy risks and the need for model evolution, and the detection process and results lack tamper-proof end-to-end evidence storage, resulting in weak traceability and legal validity.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, embodiments of the present invention provide an online detection and analysis system for traditional Chinese medicine components, comprising: a data acquisition module for acquiring multimodal data of traditional Chinese medicine samples; a fusion identification module for performing fusion analysis on the multimodal data through a built-in deep learning multi-source information fusion identification model and directly outputting structured identification conclusions; a consensus verification module for performing trusted storage through a blockchain network and using smart contracts to achieve cross-terminal data consensus verification; and a model optimization module for achieving continuous collaborative evolution of the global identification model through a network aggregation and distribution mechanism.

[0008] As a preferred embodiment of the online detection and analysis system for traditional Chinese medicine components described in this invention, the multimodal data includes spectral data, chromatographic data, mass spectrometry data, digital odor fingerprint spectrum, and taste fingerprint spectrum; the multimodal data is acquired by integrating physicochemical analysis, biomimetic olfactory sensing, and biomimetic taste sensing submodules; the structured identification conclusions include evaluation of medicinal material variety, probability of authenticity, and processing technology.

[0009] As a preferred embodiment of the online detection and analysis system for traditional Chinese medicine components described in this invention, the data acquisition module specifically includes: a physicochemical analysis unit, used to perform at least one analysis of near-infrared spectroscopy, Raman spectroscopy, or liquid chromatography to obtain chemical fingerprint data; a biomimetic olfactory sensing unit, comprising a metal oxide semiconductor sensor array and a conductive polymer sensor array, used to collect volatile components of the sample in a temperature- and humidity-controlled chamber and generate the digital odor fingerprint spectrum; and a biomimetic taste sensing unit, based on a multi-channel lipid membrane potential sensor, used to contact the sample extract and generate the digital taste fingerprint spectrum.

[0010] As a preferred embodiment of the online detection and analysis system for traditional Chinese medicine components described in this invention, the deep learning multi-source information fusion identification model built into the fusion identification module is a hierarchical multimodal fusion neural network, specifically comprising: a feature encoding sub-network: used to receive chemical spectrum data, odor fingerprint spectrum, and taste fingerprint spectrum from the data acquisition module respectively, and extract high-dimensional feature vectors through their respective convolutional neural networks or transformer encoders; a cross-modal attention fusion unit: connected to the feature encoding sub-network, used to interact with the three types of feature vectors of chemical, odor, and taste; and a structured output sub-network: connected to the cross-modal attention fusion unit, which is a multi-task learning network that outputs each component in the structured identification conclusion in parallel based on the fusion feature representation.

[0011] As a preferred embodiment of the online detection and analysis system for traditional Chinese medicine components described in this invention, the consensus verification module specifically includes: a data fingerprint generation unit, used to perform encrypted hashing operations on key fields in the structured identification conclusion and feature vectors of multimodal data used to generate the conclusion, generating an immutable, globally unique data fingerprint; a blockchain light node unit, serving as the interaction interface between the system and a permissioned blockchain network, used to bind the data fingerprint, summary information of the structured identification conclusion, terminal device identifier, and timestamp, construct a data packet for evidence storage, and broadcast it to the blockchain network; and a smart contract verification unit, wherein a specific smart contract is deployed on the blockchain network, configured to receive data packets for evidence storage of traditional Chinese medicine samples from multiple terminals for the same batch or source, extract the structured identification conclusions therein, perform similarity calculation and statistical consistency analysis, and when more than a preset number of terminal conclusions reach consensus within a preset confidence threshold, the smart contract automatically marks the conclusion as "consensus verified" and records the verification result and related data fingerprint in the immutable ledger of the blockchain, while simultaneously generating corresponding verification credentials and feeding them back to each terminal.

[0012] As a preferred embodiment of the online detection and analysis system for traditional Chinese medicine components described in this invention, the model optimization module is configured to run on a decentralized federated learning framework combined with a blockchain incentive mechanism, and specifically executes the following co-evolutionary process: Local training triggering and data selection: After the consensus verification module completes the storage and verification of the detection data, each terminal automatically triggers the local training process; the data used for training is high-credibility data samples marked as "consensus verified" by the smart contract and their corresponding multimodal raw data; Security parameter update generation: Each terminal uses selected data locally to incrementally train the deep learning multi-source information fusion and identification model in the fusion and identification module. After training, the model parameters are updated. The updated values ​​are homomorphically encrypted to generate encrypted local parameter update packages; secure aggregation and global model update: each terminal uploads the encrypted local parameter update package to an aggregation server deployed on or supervised by the blockchain network; the aggregation server, without decryption, performs a weighted average calculation on all received encrypted parameter updates using a secure multi-party computation or homomorphic encryption aggregation algorithm to generate encrypted global model update parameters; model distribution and terminal synchronization: the aggregation server distributes the encrypted global model update parameters to all participating terminals through the blockchain network; each terminal decrypts the received parameters and securely merges them with its existing local model parameters, thereby completing the collaborative evolution and upgrade of the local identification model.

[0013] As a preferred embodiment of the online detection and analysis system for traditional Chinese medicine components described in this invention, in the steps of secure aggregation and global model update, the weight of the weighted average calculation is dynamically determined by the quality score of the verification credentials attached to the "consensus verified" data samples uploaded by the smart contract to each terminal.

[0014] As a preferred embodiment of the online detection and analysis system for traditional Chinese medicine components described in this invention, in the model distribution and terminal synchronization step, the secure fusion is specifically implemented through a secure parameter averaging algorithm; after each terminal decrypts and obtains the global model update parameters, it performs a weighted average with its own local model parameters, wherein the weights assigned to the global update parameters are adaptively adjusted based on the quantity and quality score of the "consensus verified" data recently contributed by the terminal.

[0015] As a preferred embodiment of the online detection and analysis system for traditional Chinese medicine components described in this invention, the blockchain incentive mechanism is specifically manifested as a tradable digital point system. When the data contributed by the terminal is used to generate a valid global model update, the smart contract automatically issues corresponding points to its address based on its contribution. These points can be used to prioritize access to model update services, pay network fees, or exchange for other system resources.

[0016] As a preferred embodiment of the online detection and analysis system for traditional Chinese medicine components described in this invention, the encrypted global model update parameters are encapsulated as a verifiable blockchain transaction when distributed through the blockchain network; each terminal verifies the zero-knowledge proof of aggregation validity attached to the transaction and generated by the aggregation server to confirm that the global update has not been tampered with and that the calculation process is correct, and then performs decryption and fusion operations.

[0017] In a second aspect, embodiments of the present invention provide a computer device, including a memory and a processor, wherein the memory stores a computer program, and the computer program, when executed by the processor, implements any step of the online detection and analysis system for traditional Chinese medicine components as described in the first aspect of the present invention.

[0018] Thirdly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the online detection and analysis system for traditional Chinese medicine components as described in the first aspect of the present invention.

[0019] The beneficial effects of this invention are as follows: By integrating multimodal data fusion and deep learning analysis, this invention significantly improves the comprehensiveness and accuracy of traditional Chinese medicine identification, especially achieving quantitative discrimination of complex characteristics such as authenticity and processing techniques; the consensus verification mechanism based on blockchain provides tamper-proof, traceable, and cross-terminal mutually recognized credible evidence for the test results, greatly enhancing the credibility and legal validity of the results; by integrating blockchain incentives and federated learning model optimization modules, the system drives the global identification model to continuously and securely co-evolve using high-quality consensus data under the premise of strictly protecting data privacy, enabling the entire network to have autonomous adaptation and collective intelligence improvement capabilities, thereby systematically solving the core bottlenecks of information isolation, model solidification, lack of credible collaboration and evolutionary motivation in existing technologies. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a flowchart illustrating the implementation of the present invention in Example 1.

[0021] Figure 2 This is a dynamic adjustment diagram of the present invention in Example 1. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Example 1 Reference Figure 1 and Figure 2 This is the first embodiment of the present invention, which provides an online detection and analysis system for traditional Chinese medicine components, including the following steps: S1: Data acquisition module: used to acquire multimodal data of traditional Chinese medicine samples.

[0026] Preferably, the multimodal data includes spectral data, chromatographic data, mass spectrometry data, digital odor fingerprint spectrum, and taste fingerprint spectrum.

[0027] Furthermore, multimodal data is acquired by integrating physicochemical analysis, biomimetic olfactory sensing, and biomimetic gustatory sensing submodules.

[0028] Furthermore, the data acquisition module specifically includes: The physicochemical analysis unit is used to perform at least one of near-infrared spectroscopy, Raman spectroscopy, or liquid chromatography to obtain chemical fingerprint data. The biomimetic olfactory sensing unit includes a metal oxide semiconductor sensor array and a conductive polymer sensor array, which are used to collect volatile components of samples and generate digital odor fingerprint spectra in a temperature- and humidity-controlled chamber. The biomimetic taste sensing unit, based on a multi-channel lipid membrane potential sensor, is used to contact the sample extract and generate a digital taste fingerprint spectrum.

[0029] Preferably, by dynamically calculating the correlation weights between different modal features, a knowledge-level fusion of chemical spectra and biomimetic sensory information is achieved. This can automatically discover and quantify the implicit correlation between "characteristics" and chemical components in traditional experience-based identification, significantly improving the identification accuracy and interpretability of complex characteristics such as authenticity and processing techniques. Thus, subjective experience is transformed into a calculable and verifiable quantitative discrimination model.

[0030] For example, taking the identification of the authenticity of Astragalus membranaceus as an example, the data acquisition module simultaneously activates three sub-units: the physicochemical analysis unit performs near-infrared spectroscopy scanning on Astragalus membranaceus powder to obtain the characteristic absorption spectra of its polysaccharides, flavonoids and other components; the biomimetic olfactory sensing unit collects the volatile components of Astragalus membranaceus slices in a 25℃ constant temperature cavity to generate a digital odor fingerprint spectrum containing beany and sweet aromas; and the biomimetic taste sensing unit detects Astragalus membranaceus aqueous extract to generate a digital taste fingerprint spectrum with a predominantly sweet taste and a slight honey taste. The three sets of data are simultaneously timestamped and transmitted to the fusion identification module.

[0031] S2: Fusion and Identification Module: Used to perform fusion analysis on multimodal data through a built-in deep learning multi-source information fusion and identification model, and directly output structured identification conclusions.

[0032] Preferably, the structured identification conclusions include an assessment of the medicinal material variety, the probability of its authenticity, and the processing technique.

[0033] Specifically, the deep learning multi-source information fusion and identification model built into the fusion and identification module is a hierarchical multimodal fusion neural network, which includes: Feature encoding sub-network: used to receive chemical spectrum data, odor fingerprint spectrum and taste fingerprint spectrum from the data acquisition module respectively, and extract high-dimensional feature vectors through their respective convolutional neural networks or transformer encoders; Cross-modal attention fusion unit: connected to the feature encoding sub-network, used to interact with the three types of feature vectors: chemical, odor and taste; this unit dynamically models the correlation between different modal features by calculating cross-modal attention weights, and generates a unified fusion feature representation rich in contextual information; Structured output subnetwork: connected to the cross-modal attention fusion unit, which is a multi-task learning network. Based on the fusion feature representation, it outputs the various components in the structured identification conclusion in parallel, including but not limited to: the classification of medicinal materials, the probability value of authenticity, and the discrimination results of processing technology type.

[0034] Preferably, decentralized smart contracts are used to automatically verify the consistency of multi-source independent detection results and generate trusted state labels. Without the intervention of a centralized authority, a distributed, tamper-proof collaborative verification network is built. This not only greatly improves the credibility and legal validity of single detection results, but also automatically produces high-quality, high-confidence labeled data, laying a solid foundation for subsequent model evolution.

[0035] For example, after receiving the multimodal data of Astragalus membranaceus, the fusion identification module starts a hierarchical neural network: the feature encoding subnetwork uses a one-dimensional CNN to process the near-infrared spectrum and a Transformer encoder to process the odor and taste fingerprint spectrum, extracting a 128-dimensional feature vector; the cross-modal attention fusion unit calculates and finds that the attention weight of "honey flavor feature" and "specific sugar absorption peak" is as high as 0.87, and generates a fusion feature representation accordingly; the structured output subnetwork outputs the results in parallel: the variety is Astragalus membranaceus (confidence 99.2%), the probability of authenticity (from Longxi) is 94.7%, and the processing method is identified as honey-processed.

[0036] S3: Consensus Verification Module: Used for trusted storage through the blockchain network and to achieve cross-terminal data consensus verification using smart contracts.

[0037] Preferably, the consensus verification module specifically includes: The data fingerprint generation unit is used to perform encrypted hash operations on the key fields in the structured identification conclusion and the feature vectors of the multimodal data used to generate the conclusion, so as to generate an immutable, globally unique data fingerprint. The blockchain light node unit, as the interaction interface between the system and the permissioned blockchain network, is used to bind data fingerprints, summary information of structured authentication conclusions, terminal device identifiers and timestamps, construct evidence storage transaction data packets and broadcast them to the blockchain network; The smart contract verification unit is a blockchain network with a specific smart contract. This contract is configured to receive data packets of evidence-based transactions from multiple terminals for the same batch or source of traditional Chinese medicine samples, extract the structured identification conclusions, and perform similarity calculation and statistical consistency analysis. When more than a preset number of terminal conclusions reach a consensus within a preset confidence threshold, the smart contract automatically marks the conclusion as "consensus verified" and records the verification result and related data fingerprints in the immutable ledger of the blockchain. At the same time, it generates corresponding verification credentials and feeds them back to each terminal.

[0038] Preferably, the trust metric (quality score) generated by blockchain consensus is used as the weight basis for the aggregation of federated learning models. This creatively uses "data credibility" as a guiding factor for model optimization, ensuring that the evolution direction of the global model is always driven by highly credible data. This effectively defends against the pollution of low-quality or malicious data and builds an incentive-compatible ecosystem that encourages participants to actively improve data quality and detection accuracy.

[0039] For example, after five terminals distributed across the country complete the testing of the same batch of Astragalus membranaceus samples, each terminal uploads the structured data containing key conclusions such as "produced in Longxi, with a probability of authenticity > 90%" and its feature vector hash value to the blockchain. After receiving all the stored evidence, the smart contract deployed on the blockchain automatically performs a similarity analysis and finds that the probability of authenticity in all five conclusions is in the range of 92%-96% (similarity > 95%). It then automatically marks the batch of Astragalus membranaceus as "consensus verified - authentic Longxi Astragalus membranaceus" and permanently records this status and the verification certificate containing the signatures of the five terminals on the blockchain.

[0040] S4: Model Optimization Module: Used to achieve continuous collaborative evolution of the global identification model through network aggregation and distribution mechanisms.

[0041] Preferably, the model optimization module is configured to run on a decentralized federated learning framework that incorporates blockchain incentive mechanisms, and specifically executes the following co-evolutionary process: Local training triggering and data selection: After the consensus verification module completes the storage and verification of the detection data for this time, each terminal automatically triggers the local training process; The data used for training consists of high-credibility data samples that have been marked as "consensus verified" by smart contracts, along with their corresponding multimodal raw data. Security parameter update generation: Each terminal uses selected data locally to incrementally train the deep learning multi-source information fusion and identification model in the fusion and identification module. After training, the updated values ​​of the model parameters (rather than the original data or the complete model) are homomorphically encrypted to generate an encrypted local parameter update package. Secure aggregation and global model update: Each terminal uploads an encrypted local parameter update package to an aggregation server deployed on or under the supervision of a blockchain network; Without decryption, the aggregation server performs a weighted average calculation on all received encrypted parameter updates using a secure multi-party computation or homomorphic encryption aggregation algorithm to generate encrypted global model update parameters. Model distribution and terminal synchronization: The aggregation server distributes encrypted global model update parameters to all terminals participating in training via a blockchain network; Each terminal decrypts the received parameters and securely integrates them with existing local model parameters, thereby completing the collaborative evolution and upgrade of the local identification model.

[0042] Specifically, in the steps of secure aggregation and global model update, the weights of the weighted average calculation are dynamically determined by the quality score of the verification credentials attached to the "consensus verified" data samples uploaded by the smart contract to each terminal. The quality score is calculated based on the number of terminals that reached consensus during the consensus verification process, the historical verification accuracy of each terminal, and the complexity of the data itself.

[0043] Specifically, in the model distribution and terminal synchronization steps, security fusion is achieved through a secure parameter averaging algorithm; After decrypting and obtaining the global model update parameters, each terminal performs a weighted average with its own local model parameters. The weights assigned to the global update parameters are adaptively adjusted based on the quantity and quality score of the "consensus verified" data recently contributed by the terminal.

[0044] Specifically, the blockchain incentive mechanism manifests itself as a tradable digital point system. Once the data contributed by the terminal is used to generate a valid global model update, the smart contract automatically distributes corresponding points to its address based on its contribution. These points can be used to prioritize access to model update services, pay network fees, or exchange for other system resources.

[0045] Specifically, encrypted global model update parameters are encapsulated as a verifiable blockchain transaction when distributed through the blockchain network; Each terminal verifies the zero-knowledge proof of aggregation validity generated by the aggregation server attached to the transaction, confirming that the global update has not been tampered with and that the calculation process is correct, and then performs the decryption and fusion operations.

[0046] Preferably, by using zero-knowledge proof technology, the aggregation server can prove the correctness of its aggregation calculation process to each terminal without disclosing any sensitive input information. Under the premise of strictly protecting the privacy of data and model parameters of all participants, the verifiability of the collaborative calculation process is realized, eliminating the mandatory trust dependence on the aggregation server, thereby building a truly decentralized, auditable and malicious-resistant secure collaborative evolution environment.

[0047] For example, after completing the consensus verification of the above-mentioned Astragalus sample, the system automatically triggers the federated learning process: each terminal uses the "consensus verified" sample and its multimodal raw data to incrementally train the fusion identification model locally, and uploads the encrypted parameter update to the aggregation server; the server performs weighted aggregation of the parameter update based on the historical accuracy of each terminal (as a quality score) to generate a global model update; after the update is distributed through the blockchain, each terminal decrypts and integrates the new parameters, which improves the identification accuracy of the entire network model for "Longxi Astragalus honey-processed process" by 3.2%, without exposing the raw data of any terminal.

[0048] In summary, this invention significantly improves the comprehensiveness and accuracy of traditional Chinese medicine identification through multimodal data fusion and deep learning analysis, especially achieving quantitative identification of complex characteristics such as authenticity and processing techniques. The blockchain-based consensus verification mechanism provides tamper-proof, traceable, and cross-terminal mutually recognized reliable evidence for the test results, greatly enhancing the credibility and legal validity of the results. By integrating blockchain incentives and federated learning into the model optimization module, the system drives the global identification model to continuously and securely evolve collaboratively using high-quality consensus data while strictly protecting data privacy. This enables the entire network to possess autonomous adaptation and collective intelligence enhancement capabilities, thus systematically solving the core bottlenecks of information isolation, model rigidity, and lack of reliable collaboration and evolutionary impetus in existing technologies.

[0049] Example 2 is an embodiment of the present invention, which differs from the previous embodiment in that: If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0050] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0051] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0052] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0053] 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, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An online detection and analysis system for traditional Chinese medicine components, characterized in that: include: Data acquisition module: used to acquire multimodal data of traditional Chinese medicine samples; Fusion and identification module: It is used to perform fusion analysis on multimodal data through the built-in deep learning multi-source information fusion and identification model, and directly output structured identification conclusions; Consensus verification module: used for trusted storage through the blockchain network and to realize cross-terminal data consensus verification using smart contracts; Model optimization module: Used to achieve continuous collaborative evolution of the global identification model through network aggregation and distribution mechanisms.

2. The online detection and analysis system for traditional Chinese medicine components as described in claim 1, characterized in that: The multimodal data includes spectral data, chromatographic data, mass spectrometry data, digital odor fingerprint spectrum, and taste fingerprint spectrum; The multimodal data is acquired by integrating physicochemical analysis, biomimetic olfactory sensing, and biomimetic gustatory sensing submodules; The structured identification conclusions include assessments of medicinal material varieties, probability of origin, and processing techniques.

3. The online detection and analysis system for traditional Chinese medicine components as described in claim 2, characterized in that: The data acquisition module specifically includes: The physicochemical analysis unit is used to perform at least one of near-infrared spectroscopy, Raman spectroscopy, or liquid chromatography to obtain chemical fingerprint data. The biomimetic olfactory sensing unit includes a metal oxide semiconductor sensor array and a conductive polymer sensor array, used to collect volatile components of a sample in a temperature- and humidity-controlled chamber and generate the digital odor fingerprint spectrum. The biomimetic taste sensing unit, based on a multi-channel lipid membrane potential sensor, is used to contact the sample extract and generate the digital taste fingerprint spectrum.

4. The online detection and analysis system for traditional Chinese medicine components as described in claim 1, characterized in that: The deep learning multi-source information fusion and identification model built into the fusion and identification module is a hierarchical multimodal fusion neural network, which specifically includes: Feature encoding sub-network: used to receive chemical spectrum data, odor fingerprint spectrum and taste fingerprint spectrum from the data acquisition module respectively, and extract high-dimensional feature vectors through their respective convolutional neural networks or converter encoders; Cross-modal attention fusion unit: connected to the feature encoding sub-network, used to interact with the three types of feature vectors: chemical, odor, and taste; Structured output sub-network: connected to the cross-modal attention fusion unit, which is a multi-task learning network that outputs the components in the structured discrimination conclusion in parallel based on the fusion feature representation.

5. The online detection and analysis system for traditional Chinese medicine components as described in claim 1, characterized in that: The consensus verification module specifically includes: The data fingerprint generation unit is used to perform encrypted hash operations on the key fields in the structured identification conclusion and the feature vectors of the multimodal data used to generate the conclusion, so as to generate an immutable, globally unique data fingerprint. The blockchain light node unit, as the interaction interface between the system and the permissioned blockchain network, is used to bind the data fingerprint, the summary information of the structured authentication conclusion, the terminal device identifier and the timestamp, construct the evidence storage transaction data packet and broadcast it to the blockchain network; The smart contract verification unit is configured to receive data packets from multiple terminals regarding the same batch or source of traditional Chinese medicine samples, extract the structured identification conclusions, perform similarity calculations and statistical consistency analysis, and automatically mark the conclusion as "consensus verified" when more than a preset number of terminal conclusions reach a consensus within a preset confidence threshold. The smart contract also records the verification results and related data fingerprints in the immutable ledger of the blockchain, and generates corresponding verification credentials to feed back to each terminal.

6. The online detection and analysis system for traditional Chinese medicine components as described in claim 5, characterized in that: The model optimization module is configured to run on a decentralized federated learning framework that incorporates a blockchain incentive mechanism, and specifically executes the following co-evolutionary process: Local training triggering and data selection: After the consensus verification module completes the storage and verification of the detection data for this time, each terminal automatically triggers the local training process; The data used for training consists of high-confidence data samples that have been marked as "consensus verified" by the smart contract and their corresponding multimodal raw data; Security parameter update generation: Each terminal uses selected data locally to incrementally train the deep learning multi-source information fusion and identification model in the fusion and identification module. After training is completed, the updated values ​​of the model parameters are homomorphically encrypted to generate an encrypted local parameter update package. Secure aggregation and global model update: Each terminal uploads the encrypted local parameter update package to an aggregation server deployed on or under the supervision of the blockchain network; Without decryption, the aggregation server performs a weighted average calculation on all received encrypted parameter updates using a secure multi-party computation or homomorphic encryption aggregation algorithm to generate encrypted global model update parameters. Model distribution and terminal synchronization: The aggregation server distributes encrypted global model update parameters to all terminals participating in training through the blockchain network; Each terminal decrypts the received parameters and securely integrates them with existing local model parameters, thereby completing the collaborative evolution and upgrade of the local identification model.

7. The online detection and analysis system for traditional Chinese medicine components as described in claim 6, characterized in that: In the secure aggregation and global model update steps, the weights of the weighted average calculation are dynamically determined by the quality score of the verification credentials attached to the "consensus verified" data samples uploaded by the smart contract to each terminal.

8. The online detection and analysis system for traditional Chinese medicine components as described in claim 6, characterized in that: In the model distribution and terminal synchronization steps, the security fusion is specifically implemented through a secure parameter averaging algorithm; After decrypting and obtaining the global model update parameters, each terminal performs a weighted average with its own local model parameters. The weights assigned to the global update parameters are adaptively adjusted based on the quantity and quality score of the "consensus verified" data recently contributed by the terminal.

9. The online detection and analysis system for traditional Chinese medicine components as described in claim 6, characterized in that: The blockchain incentive mechanism is specifically manifested as a tradable digital point; Once the data contributed by the terminal is used to generate a valid global model update, the smart contract automatically distributes corresponding points to its address based on its contribution. These points can be used to prioritize access to model update services, pay network fees, or exchange for other system resources.

10. The online detection and analysis system for traditional Chinese medicine components as described in claim 6, characterized in that: The encrypted global model update parameters are encapsulated as a verifiable blockchain transaction when distributed through the blockchain network. Each terminal verifies the zero-knowledge proof of aggregation validity generated by the aggregation server attached to the transaction, confirming that the global update has not been tampered with and that the calculation process is correct, and then performs the decryption and fusion operations.

Images