Automobile part multi-source fusion tracing method and system based on distributed block chain

Abstract

The invention discloses an automobile part multi-source fusion tracing method and system based on a distributed block chain. Firstly, a block chain network and a tracing platform are constructed; a microscopic visual feature image of the surface of the automobile part is collected, the texture type is judged, features are extracted, electromagnetic field intensity distribution features are extracted in combination with an RFID tag, and electromagnetic fingerprints are recognized through a neural network; secondly, multi-source feature data and production and transportation information are bound with unique identifiers and then stored in a distributed block chain network, and a five-layer Merkel tree storage structure is adopted to ensure that the data cannot be tampered; and finally, constructing a tracing query module of a three-layer architecture, and automatically managing data writing and query through an intelligent contract. According to the invention, through combination of multi-feature fusion and the block chain technology, the problems of information island and weak traceability of a traditional traceability system are solved, credible traceability and authenticity identification of automobile parts in the whole life cycle are realized, and the supply chain data security and query efficiency are improved.

Description

Technical Field

[0001] The present invention belongs to the field of blockchain technology application and Internet of Things traceability, and in particular relates to a multi-source fusion traceability method and system for automotive parts based on distributed blockchain. Background Art

[0002] The counterfeiting and passing off of inferior auto parts as genuine ones has always been a serious problem. Some unscrupulous vendors use low-quality materials or craftsmanship to produce counterfeit auto parts that pass themselves off as imported brands. While their appearance and logos may closely resemble genuine products, their quality and performance often fall short of genuine product standards, making it difficult to meet safety and reliability requirements. These counterfeit and inferior auto parts not only harm consumers and hinder the healthy and orderly development of the entire industry and market, but also result in significant losses in import tariffs and damage my country's image and international standing. Given the widespread counterfeiting and passing-off of inferior auto parts, fully leveraging existing feature extraction and recognition technologies for auto parts is crucial for authenticity verification and combating counterfeiting and smuggling in the auto parts industry.

[0003] However, relying solely on a unique ID for each item is no longer sufficient to solve the traceability problem for auto parts. "Object fingerprint recognition technology" can achieve highly accurate identification in a fraction of the time by taking a photo of an auto part and comparing it with the product's texture image data, which has been previously uploaded and registered in the cloud by the brand. The metal, fiber, glass, leather, or plastic used in auto parts all have unique textures that are invisible to the naked eye, known as object fingerprints.

[0004] While traceability and identification technologies based on texture fingerprint recognition already exist, manual involvement in identification is still significant, and the primary outcome depends on the judgment of experienced appraisers. This project will investigate automotive component feature extraction and intelligent comparison technologies based on computer vision and artificial intelligence. This technology will perform image processing such as stretching and transforming captured photos of automotive components, then automatically extract two-dimensional and three-dimensional features. This technology will then utilize neural networks to further identify and match these features, enabling rapid identification and automated comparison. This will further reduce uncertainty in automotive component identification results and provide effective support for reliable traceability of automotive components.

[0005] The electromagnetic fingerprint information hidden in automotive parts can be collected using specific sensors. These sensors, typically including magnetoresistive sensors and Hall effect sensors, detect the component's response to a magnetic field and convert it into an electrical signal. These signals are then received and digitized by a data collector. By analyzing the pre-processed signals, various features reflecting the electromagnetic properties of the automotive component, such as frequency, amplitude, phase, and waveform, can be extracted. These features can reveal the true characteristics and potential differences of the component.

[0006] The blockchain's transparency, security, and uniqueness make it a natural fit for reliable traceability. Auto parts traceability collects key product information and stores it on the blockchain. By maintaining a permanent, immutable, timestamp-based blockchain data network, the uniqueness and authenticity of this key information can be established. Multi-source data fusion technology is also used to combine heterogeneous traceability data from various links and stages through joint analysis of auto parts' traceability tags, hidden electromagnetic fingerprints, and product characteristics. This combined data generates a historical record of auto parts traceability, meeting all requirements for reliable traceability. This technology can also be used to perform similarity analysis on the appearance and structure of auto parts, determining whether any auto parts infringe intellectual property rights or other issues through similarity analysis, effectively protecting the healthy development of the auto parts industry. Summary of the Invention

[0007] In response to the problems of information islands and easy data tampering in the existing technology of automobile parts traceability, the present invention proposes a multi-source fusion traceability method and system for automobile parts based on distributed blockchain. By integrating microscopic visual features with RFID electromagnetic fingerprints and production information, transportation information, etc., and combining blockchain distributed storage and smart contract technology, the method and system can achieve reliable recording, tamper-proof traceability and automated authenticity identification of parts throughout their life cycle, effectively improving supply chain management efficiency.

[0008] The technical solution to achieve the purpose of the present invention is: A multi-source fusion traceability method for automobile parts based on distributed blockchain, characterized by comprising the following steps: Step S1: Build blockchain network module: Each business entity in the automotive parts supply chain is added to the blockchain as a node. The business entities include upstream suppliers, manufacturers, downstream distributors, retailers and customers. An authentication center is set up in the blockchain network, with each supplier, manufacturer, and seller acting as an independent authentication center to authenticate the authenticity of product information and the rationality of decisions in the automotive parts supply chain; Construct a multi-source data fusion analysis module as a node to join the blockchain network module. The multi-source data fusion analysis module is used to receive information collection requirements transmitted by various business entities in the supply chain, determine the data set according to the requirements, perform data collection and processing, and then return the processing results to each business entity in the supply chain; Step S2: constructing a multi-source fusion traceability platform module for automobile parts, wherein the multi-source fusion traceability platform module is used to record information of each link in the automobile parts supply chain and use timestamps to record the time sequence; Step S3: Multi-source data collection to form unique identity information of auto parts: Collect microscopic visual feature images of the surfaces of automotive parts and components, and establish a dataset of microscopic visual feature images of the surfaces of various automotive parts and components; Using a multi-antenna array, a radio frequency tag reader collects the electromagnetic field intensity distribution characteristics of the hidden electromagnetic fingerprint RFID tag on the outer packaging of automotive parts, and uses a neural network algorithm to perform feature analysis and classification identification on the extracted electromagnetic fingerprint data; Collect production information, transportation information, inspection information, and import and export information; Step S4, multi-source data fusion: sending the data collected in step S3 to the multi-source data fusion analysis module, and using the multi-source data fusion analysis module to process and fuse the data; Step S5: Data on-chain storage: The fused data is written to the blockchain through a smart contract module pre-deployed on the blockchain. The smart contract is a computer protocol that automatically executes the contract terms and is used to automatically verify and store data. Each transaction record is packaged into a block, which contains the hash value of the current block, the hash value of the previous block, and timestamp information. Step S6, traceability query: The user inputs parts-related information through the automobile parts multi-source fusion traceability platform module, and the automobile parts multi-source fusion traceability platform module searches for matching traceability data on the blockchain according to the query rules in the smart contract module.

[0009] Furthermore, the blockchain network in step S1 adopts a distributed storage structure, which is used to store different types of basic data. Different types of data are mapped to the block bodies of corresponding blocks in the blockchain. The distributed storage data structure consists of the following five parts: Data layer: refers to supply chain management related data; Classification Merkle Tree Layer: Data of different categories form different Merkle trees, and the Merkle trees do not intersect with each other. This layer is used to store Merkle trees; Classification Merkle tree root layer: This layer stores the roots of different Merkle trees; Directory layer: This layer constructs a directory based on the basic data corresponding to different classification Merkle roots, and calculates a Merkle root from the classification Merkle roots. The number of classification Merkle roots can be selected according to actual needs. Block header layer: All Merkle tree roots in the directory layer are calculated to form the total Merkle tree root. The block header will contain this tree root, including but not limited to the block header hash value of the previous block, the version number of this block, consensus authentication parameters, and timestamp.

[0010] Furthermore, the multi-source fusion traceability platform module for automotive parts in step S2 adopts a three-layer architecture design, namely, a data application layer, a data management layer, and a data collection layer; wherein, the data application layer provides traceability information query services for consumers and suppliers, the data management layer is responsible for managing the storage, processing, and maintenance of data required for various business functions, and the data collection layer is responsible for collecting and acquiring relevant data during the production process of automotive parts; and in the platform, each business entity only has the authority to distribute the storage blockchain and stores the data in a distributed manner.

[0011] Furthermore, in step S3, the collected microscopic visual feature image of the surface of the automotive parts material is preprocessed, including image segmentation, obtaining a directional map, enhancement and filtering, binarization and refinement; and the texture type is judged according to its two-dimensional frequency domain characteristics, and the corresponding algorithm is selected to extract and analyze the key attribute feature points of the texture features; the texture types include continuous, discontinuous, and contour types. For continuous textures, the Poincare index algorithm is selected, for discontinuous textures, the gray-level co-occurrence matrix GLCM algorithm is selected, and for contour textures, the Freeman chain code and its derivative feature extraction algorithm are used; the extracted texture features are analyzed and matched using the deep learning framework CNN.

[0012] Furthermore, in step S3, the step of collecting the electromagnetic field intensity distribution characteristics of the hidden electromagnetic fingerprint RFID tag includes: Step S3-1: Perform IQ demodulation on the RFID tag signal received by the RFID tag reader to obtain I and Q signals, and obtain the amplitude signal after modulus calculation. ,in, , is the sampling point; Step S3-2: Perform clustering to obtain the center point vector corresponding to code element 0 and 1 ,in is the cluster center corresponding to code element 0, is the cluster center point corresponding to code element 1, and the clustering function is expressed as: ,in, is the clustering operation function; Step S3-3: Determine based on clustering of silent period signal and EPC signal and , the expected signal is obtained by Euclidean distance judgment ,in, , is the Euclidean distance decision function; Step S3-4: Amplitude signal Normalize to remove amplitude differences: , in, is a standard amplitude signal; Step S3-5: Standard amplitude signal Removing the desired signal To calculate the noise signal : , in, is a noise signal; Step S3-6: Expected signal , standard amplitude signal , noise signal and the original signal By calculating the cross spectral entropy and related features through pairwise combination, the cross spectral entropy features are extracted, and the spectral entropy Defined as: , in, For signal In frequency The energy proportion probability density function at : , The cross power spectral density is expressed based on the Wiener-Schinchin theorem as: , in, For signal and The cross power spectrum reflects the correlation between signals in the frequency domain. for The conjugate of represents the Fourier transform, For signal and The cross-correlation function is used to evaluate the cross-correlation between two signals. The similarity of is the delay variable; From the above formula, the cross spectral entropy can be obtained as: , in, For signal and The cross spectral entropy of is used as the feature of RFID tags for automotive parts; is the energy proportion probability density function of the cross power spectrum; At the same time, the introduction of the maximum , maximum phase , maximum frequency , and cross spectral entropy Together they serve as the characteristics of RFID tags for automotive parts; their expressions are: .

[0013] Furthermore, in step S3, a neural network algorithm is used to process the multi-source fusion features of the components extracted in step S3-6, specifically including: , maximum phase , maximum frequency , and cross spectral entropy As the input of the neural network, the network is trained by introducing a back-propagation algorithm optimized by the momentum term. After the training is completed, the network is used to perform classification inference on the input fusion features, output the authenticity determination results of the parts and the corresponding traceability information, and write the results into the distributed blockchain to complete the evidence storage.

[0014] Furthermore, in step S5, the smart contract module completes decision-making activities and performs distributed accounting through decentralized consensus authentication, so that the smart contract and product data are traceable; in the storage link, the user calls the smart contract to trace the product information on the chain; the uploaded information will be recorded in the blockchain ledger and a transaction hash address will be returned as the unique identifier of the information in the blockchain.

[0015] Furthermore, each business entity only has the authority to store distributed blockchain data. They can only perform data storage operations after encrypting their identity data with an encryption key. They do not have the right to store business-related data that has a mapping relationship with the blockchain. The encryption key is collaboratively managed and backed up by multiple authoritative automotive parts industry organizations. When verifying the authenticity of automotive parts, the encryption key provided by the supply chain entity and the official backup key are used to decrypt the relevant product data respectively, and the consistency of the two decryption results is compared to determine whether the supply chain entity has tampered with the product data key.

[0016] Furthermore, the automotive parts supply chain is built on an industry-specific network or the Internet, and requires that only business entities in the supply chain be authorized to join the blockchain system.

[0017] A distributed blockchain-based multi-source fusion traceability system for automobile parts, applied to a distributed blockchain-based multi-source fusion traceability method for automobile parts, is characterized by including: Blockchain network module: used to build a blockchain network, adding certification centers, suppliers, manufacturers, sellers, and multi-source data fusion analysis centers as nodes to achieve information sharing and supervision; Auto parts multi-source fusion traceability platform module: used to build an auto parts multi-source fusion traceability platform; Multi-source data acquisition module: used to collect component information at all stages of the automotive parts process, including production time, batch, raw material source, processing technology, transportation route, and sales records; Multi-source data fusion analysis module: used to process and fuse the collected data to form complete traceability information; Data on-chain module: used to write the integrated data into the blockchain through smart contracts to ensure that the data cannot be tampered with and is traceable; Traceability query module: used to receive user query requests, search for matching traceability data on the blockchain according to the query rules in the smart contract, and display it to the user; Smart contract module: used to achieve traceability and automatic collection of automotive parts supply chain information, and complete decision-making activities and conduct distributed accounting through decentralized consensus authentication based on the automatic operation of the blockchain.

[0018] Compared with the prior art, the present invention adopts the above technical solution and has the following beneficial effects: (1) By integrating the dual physical characteristics of the material surface microscopic visual features and the RFID electromagnetic fingerprint, a composite identification mechanism of physical fingerprint and digital identity is formed. The microscopic visual features are adapted to different algorithms based on texture type to extract key features, and the electromagnetic fingerprint is analyzed through a neural network to analyze the electromagnetic field intensity distribution characteristics. The two are combined with production and transportation information to form an unreplicable component identity, which solves the problem of traditional single labels being easy to forge and imitate from a technical principle.

[0019] (2) Using a blockchain distributed network architecture, data from all links of the supply chain is stored through a five-layer Merkle tree. Each block contains the hash value and timestamp of the previous block, building a chain encryption structure. This ensures that data cannot be tampered with once it is on the chain. At the same time, through a collaborative authentication mechanism in which each business entity acts as an independent authentication center, it solves the information island and data tampering risks in the traditional traceability system and establishes a trusted data chain for the entire supply chain process.

[0020] (3) Through the pre-deployed smart contract module, the automatic verification, storage and distributed accounting of data on the chain are realized, and the real-time chain-up of production, transportation, inspection and other information can be completed without human intervention. In the traceability link, the smart contract supports rapid retrieval and cross-spectral entropy feature comparison based on unique identifiers. Combined with the collaborative work of the three-tier architecture platform, it realizes efficient query and automated authenticity determination, significantly reducing the time cost of traditional manual verification.

[0021] (4) Through blockchain timestamp technology and distributed ledger, the entire process of parts from production to consumption is recorded in sequence, supporting two-way queries for forward tracing and reverse recall. The three-tier architecture platform further integrates data collection, management, and application services, providing a unified information sharing and supervision interface for all business entities, solving the problems of cross-link data asynchrony and difficulty in defining responsibilities in traditional systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a flow chart of the multi-source fusion traceability method for automotive parts based on distributed blockchain proposed by the present invention; Figure 2 This is a diagram of the automotive parts supply chain architecture based on distributed blockchain; Figure 3 This is the overall architecture diagram of the multi-source integrated traceability platform for automotive parts; Figure 4 This is a diagram of the distributed storage data structure of blocks in the blockchain-based supply chain management system; Figure 5 A diagram of a data storage method for authenticating the automotive parts supply chain; Figure 6 This is a diagram showing the scanning process of the genuine automotive ceramic brake pad products under a microscope in Example 1; Figure 7 This is the microscopic feature image of a real automotive ceramic brake pad product in Example 1; Figure 8 This is the microscopic characteristic image of the fake automotive ceramic brake pad product in Example 1; Figure 9 This is the image enhancement result of the microscopic features of a real automotive ceramic brake pad product in Example 1; Figure 10 This is the image enhancement result of the microscopic features of the fake automotive ceramic brake pad product in Example 1; Figure 11 This is a physical picture of the radio frequency identification test equipment in Example 2; Figure 12 This is a diagram of the 3×3 multi-label rectangular array experimental setup in Example 2. DETAILED DESCRIPTION

[0023] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0024] A multi-source fusion traceability method for automobile parts based on distributed blockchain, the process is as follows Figure 1 As shown, it is characterized in that it includes the following steps: Step S1: Build blockchain network module: Each business entity in the automotive parts supply chain is added to the blockchain as a node. The business entities include upstream suppliers, manufacturers, downstream distributors, retailers and customers. An authentication center is set up in the blockchain network, such as Figure 2 As shown, Figure 2 In the blockchain, AC stands for Certification Center. Each supplier, manufacturer, and distributor in the entire automotive parts supply chain acts as an independent certification center to authenticate the authenticity of product information and the rationality of decision-making in the automotive parts supply chain. Distributed storage of automotive parts supply chain information data in the blockchain system allows authorized access to this data by each business entity, effectively reducing the probability of auto parts being replaced by counterfeit and inferior products, or being sold as "inferior goods"; Construct a multi-source data fusion analysis module as a node to join the blockchain network module. The multi-source data fusion analysis module is used to receive information collection requirements transmitted by various business entities in the supply chain, determine the data set according to the requirements, perform data collection and processing, and then return the processing results to each business entity in the supply chain; Step S2: Construct a multi-source fusion traceability platform module for auto parts, such as Figure 3 As shown, the automotive parts multi-source fusion traceability platform module is used to record information of each link in the automotive parts supply chain and use timestamps to record the time sequence; Step S3: Multi-source data collection to form unique identity information of auto parts: Collect microscopic visual feature images of the surfaces of automotive parts and components, and establish a dataset of microscopic visual feature images of the surfaces of various automotive parts and components; Using a multi-antenna array, a radio frequency tag reader collects the electromagnetic field intensity distribution characteristics of the hidden electromagnetic fingerprint RFID tag on the outer packaging of automotive parts, and uses a neural network algorithm to perform feature analysis and classification identification on the extracted electromagnetic fingerprint data; Collect production information, transportation information, inspection information, and import and export information; Step S4, multi-source data fusion: sending the data collected in step S3 to the multi-source data fusion analysis module, and using the multi-source data fusion analysis module to process and fuse the data; Step S5: Data on-chain storage: The fused data is written to the blockchain through a smart contract module pre-deployed on the blockchain. The smart contract is a computer protocol that automatically executes the contract terms and is used to automatically verify and store data. Each transaction record is packaged into a block, which contains the hash value of the current block, the hash value of the previous block, and timestamp information. Step S6, traceability query: The user inputs parts-related information through the automobile parts multi-source fusion traceability platform module, and the automobile parts multi-source fusion traceability platform module searches for matching traceability data on the blockchain according to the query rules in the smart contract module.

[0025] Furthermore, the blockchain network in step S1 adopts a distributed storage structure, such as Figure 3 As shown in FIG, the distributed storage structure is used to store different types of basic data. Different types of data are mapped to the block bodies of corresponding blocks in the blockchain. The distributed storage data structure consists of the following five parts: Data layer: refers to supply chain management related data; Classification Merkle Tree Layer: Data of different categories form different Merkle trees, and the Merkle trees do not intersect with each other. This layer is used to store Merkle trees; Classification Merkle tree root layer: This layer stores the roots of different Merkle trees; Directory layer: This layer constructs a directory based on the basic data corresponding to different classification Merkle roots, and calculates a Merkle root from the classification Merkle roots. The number of classification Merkle roots can be selected according to actual needs. Block header layer: All Merkle tree roots in the directory layer are calculated to form the total Merkle tree root. The block header will contain this tree root, including but not limited to the block header hash value of the previous block, the version number of this block, consensus authentication parameters, and timestamp.

[0026] Furthermore, the multi-source fusion traceability platform module for automotive parts in step S2 adopts a three-layer architecture design, namely, a data application layer, a data management layer, and a data collection layer; wherein, the data application layer provides traceability information query services for consumers and suppliers, the data management layer is responsible for managing the storage, processing, and maintenance of data required for various business functions, and the data collection layer is responsible for collecting and acquiring relevant data during the production process of automotive parts; and in the platform, each business entity only has the authority to distribute the storage blockchain and stores the data in a distributed manner.

[0027] Furthermore, in step S3, the collected microscopic visual feature image of the surface of the automotive parts material is preprocessed, including image segmentation, obtaining a directional map, enhancement and filtering, binarization and refinement; and the texture type is judged according to its two-dimensional frequency domain characteristics, and the corresponding algorithm is selected to extract and analyze the key attribute feature points of the texture features; the texture types include continuous, discontinuous, and contour types. For continuous textures, the Poincare index algorithm is selected, for discontinuous textures, the gray-level co-occurrence matrix GLCM algorithm is selected, and for contour textures, the Freeman chain code and its derivative feature extraction algorithm are used; the extracted texture features are analyzed and matched using the deep learning framework CNN.

[0028] Microscopic feature images of automobile parts It is stationary in small local regions and can be described by the following model: , Where, , For exist The local mean and variance in a small neighborhood of a point is a Gaussian process with zero mean and unit variance.

[0029] For a pixel Size Neighborhood of: , Where, 、 They are hour Second moments and second-order cumulants of a point.

[0030] , Where, 、 They are hour The third-order moments and third-order cumulants of a point.

[0031] when When taking 8×8 blocks from the foreground and background areas of the microscopic feature image of the automotive parts, and comparing the grayscale histograms, we found that the grayscale of the foreground area of ​​the microscopic feature image of the automotive parts is mainly distributed between black (grayscale value 0) and white (grayscale value 255), showing a regular distribution with a large variance. On the other hand, the grayscale of the background area of ​​the microscopic feature image of the automotive parts is mainly composed of noise and can be approximately Gaussian distributed. By calculating the third-order cumulant, we set an appropriate threshold based on the actual image grayscale value parameters. T ,when When , the block is the background area of ​​the microscopic feature image, and this part of the area is cut off.

[0032] Then we find the directional map of the microscopic feature image of the automobile parts. Representing the microscopic features of automotive parts The grayscale value at . The image is divided into The small block size is set to , where the values ​​of M and N make it ; Calculate each pixel in the large block based on the sobel operator Gradients on the x-axis and y-axis and .

[0033] , Calculate pixel points The direction of the centered block .

[0034] , When the big block or When the number of zeros reaches a certain level, such as more than 80% of the total number of pixels in the block, Should be set directly to 0 or . Large block direction As The direction of the small block is saved and the direction of the next block is calculated in a loop.

[0035] Then, the microscopic feature images of automobile parts are enhanced and filtered. A two-dimensional Gabor filter with good direction and frequency selectivity is selected to remove image noise and retain the texture structure of the microscopic image. Before filtering the microscopic feature images of automobile parts, the areas with obvious material property characteristics are normalized, and their mean and variance are fixed to constants. Normalization can eliminate sensor noise and grayscale value differences caused by different pressures. Here we use a block processing method for the microscopic feature images of automobile parts. Definition Pixels The gray value at , Block The mean and variance of It's a pixel The gray value after normalization. All pixels in , the normalized image is defined as: , , are the expected mean and variance respectively. Normalization is performed uniformly on all pixels, so it does not change the texture structure of the microscopic feature image of automobile parts.

[0036] Selecting microscopic feature images of automobile parts The mathematical expression of the two-dimensional Gabor function in the spatial domain is as follows: , in, Indicates the wavelength, its value is usually greater than or equal to 2, in pixels, and cannot be greater than one-fifth of the input image size. Indicates the direction, with a value from 0 to 360 degrees, which specifies the direction of the Gabor function parallel stripes. Indicates the aspect ratio (spatial aspect ratio), which determines the ellipticity of the Gabor function curve shape. Usually the value is 0.5. When , the curve shape is circular; when When the stripes are parallel to each other, the shape of the stripes is continuously stretched. It represents the standard deviation of the Gaussian part. Its value cannot be given directly and is restricted by the center frequency of the filter. x ,y Represents the spatial position coordinates of image pixels.

[0037] Then the feature map of the microscopic feature image of automobile parts after image filtering and enhancement is expressed as: , Because a specific component in the frequency domain of a microscopic feature image of an automotive part contains the spatial distribution of the image, the present invention, based on the Fourier spectrum's ability to characterize texture features, determines and classifies the texture features of the microscopic feature image of an automotive part as either continuous or discontinuous in the two-dimensional frequency domain. Fourier spectrum texture analysis converts the image from the spatial domain to the frequency domain via Fourier transform, thereby obtaining the frequency distribution of the microscopic feature image of the automotive part. After performing feature texture classification, an automatic matching algorithm is used to match the appropriate texture feature extraction algorithm to the different texture types to extract and analyze key attribute points.

[0038] Transform the microscopic feature images of automobile parts from the spatial domain to the frequency domain. If the texture features of the microscopic feature image of an automobile part belong to two-dimensional discrete data, the Fourier transform process is as follows: , The microscopic feature image of automobile parts is transformed from the spatial domain to the frequency domain. If the texture features of the microscopic feature image of automobile parts belong to two-dimensional continuous data, the process of Fourier transform is: , Furthermore, in step S3, the step of collecting the electromagnetic field intensity distribution characteristics of the hidden electromagnetic fingerprint RFID tag includes: Step S3-1: Perform IQ demodulation on the RFID tag signal received by the RFID tag reader to obtain I and Q signals, and obtain the amplitude signal after modulus calculation. ,in, , is the sampling point; Step S3-2: Perform clustering to obtain the center point vector corresponding to code element 0 and 1 ,in is the cluster center corresponding to code element 0, is the cluster center point corresponding to code element 1, and the clustering function is expressed as: ,in, is the clustering operation function; Step S3-3: Determine based on clustering of silent period signal and EPC signal and , the expected signal is obtained by Euclidean distance judgment ,in, , is the Euclidean distance decision function; Step S3-4: Amplitude signal Normalize to remove amplitude differences: , in, is a standard amplitude signal; Step S3-5: Standard amplitude signal Removing the desired signal To calculate the noise signal : , in, is a noise signal; Step S3-6: Expected signal , standard amplitude signal , noise signal and the original signal By calculating the cross spectral entropy and related features through pairwise combination, the cross spectral entropy features are extracted, and the spectral entropy Defined as: , in, For signal In frequency The energy proportion probability density function at : , The cross power spectral density is expressed based on the Wiener-Schinchin theorem as: , in, For signal and The cross power spectrum reflects the correlation between signals in the frequency domain. for The conjugate of represents the Fourier transform, For signal and The cross-correlation function is used to evaluate the cross-correlation between two signals. The similarity of is the delay variable; From the above formula, the cross spectral entropy can be obtained as: , in, For signal and The cross spectral entropy of is used as the feature of RFID tags for automotive parts; is the energy proportion probability density function of the cross power spectrum; At the same time, the introduction of the maximum , maximum phase , maximum frequency , and cross spectral entropy Together they serve as the characteristics of RFID tags for automotive parts; their expressions are: .

[0039] Furthermore, in step S3, a neural network algorithm is used to process the multi-source fusion features of the components extracted in step S3-6, specifically including: , maximum phase , maximum frequency , and cross spectral entropy As the input of the neural network, the network is trained by introducing a back-propagation algorithm optimized by the momentum term. After the training is completed, the network is used to perform classification inference on the input fusion features, output the authenticity determination results of the parts and the corresponding traceability information, and write the results into the distributed blockchain to complete the evidence storage.

[0040] Furthermore, in step S5, the smart contract module completes decision-making activities and performs distributed accounting through decentralized consensus authentication, so that the smart contract and product data are traceable; in the storage link, the user calls the smart contract to trace the product information on the chain; the uploaded information will be recorded in the blockchain ledger and a transaction hash address will be returned as the unique identifier of the information in the blockchain.

[0041] Furthermore, each business entity only has the authority to store distributed blockchains. It can only perform data storage operations after encrypting its identity data with an encryption key, and has no right to store business-related data that has a mapping relationship with the blockchain. The encryption key is collaboratively managed and backed up by multiple authoritative automotive parts industry organizations, such as Figure 5 As shown; When verifying the authenticity of automotive parts, the encryption key provided by the supply chain entity and the official backup key are used to decrypt the relevant product data respectively, and the consistency of the two decryption results is compared to determine whether the supply chain entity has tampered with the product data key.

[0042] Furthermore, the automotive parts supply chain is built on an industry-specific network or the Internet, and requires that only business entities in the supply chain be authorized to join the blockchain system.

[0043] A distributed blockchain-based multi-source fusion traceability system for automobile parts, applied to a distributed blockchain-based multi-source fusion traceability method for automobile parts, is characterized by including: Blockchain network module: used to build a blockchain network, adding certification centers, suppliers, manufacturers, sellers, and multi-source data fusion analysis centers as nodes to achieve information sharing and supervision; Auto parts multi-source fusion traceability platform module: used to build an auto parts multi-source fusion traceability platform; Multi-source data acquisition module: used to collect component information at all stages of the automotive parts process, including production time, batch, raw material source, processing technology, transportation route, and sales records; Multi-source data fusion analysis module: used to process and fuse the collected data to form complete traceability information; Data on-chain module: used to write the integrated data into the blockchain through smart contracts to ensure that the data cannot be tampered with and is traceable; Traceability query module: used to receive user query requests, search for matching traceability data on the blockchain according to the query rules in the smart contract, and display it to the user; Smart contract module: used to achieve traceability and automatic collection of automotive parts supply chain information, and complete decision-making activities and conduct distributed accounting through decentralized consensus authentication based on the automatic operation of the blockchain.

[0044] [Example 1] Example 1 demonstrates the experimental process and results of collecting microscopic visual feature images of automotive component surfaces in step S3. Ceramic brake pads, specifically those manufactured by different manufacturers for the same vehicle model, were selected as the research subjects. Microscopic visual feature images of their surfaces were collected as a dataset of automotive component surface microscopic visual feature images to verify the effectiveness of this method. The dataset contains 100 microscopic visual feature images of automotive ceramic brake pads, 70 of which were selected as training samples, 10 as validation samples, and 20 as test samples.

[0045] The energy, contrast, correlation, entropy, etc. of the gray-level co-occurrence matrix are used as feature parameters to describe the texture information of the microscopic visual images of automobile parts.

[0046] The image processing of the microscopic visual features of the surface of automotive parts materials is implemented in Matlab R2018b. The CNN neural network model for microscopic visual feature recognition of automotive parts uses a 64-bit Ubuntu 18.04 operating system and the environment configuration is Python 3.8, PyTorch 1.4.0 combined with the Pycharm compiler.

[0047] like Figure 6 As shown in the figure, the image processing of the microscopic visual features of the surface of automotive parts materials uses authentic and fake automotive ceramic brake pad experimental materials and is scanned using a KC-X1000 series laser spectral confocal microscope.

[0048] The experimental results are as follows Figure 7 、 Figure 8 As shown in the figure, the material difference between the genuine and fake automotive ceramic brake pads can be seen by comparing the microscopic feature images of the two. The images are then preprocessed, and the processing results are shown in the figure. Figure 9 、 Figure 10 shown.

[0049] It can be seen that the feature extraction based on the hybrid algorithm of microscopic visual feature extraction and neural network can clearly see the difference between genuine and fake parts.

[0050] [Example 2] Example 2 shows the experimental process and experimental results of collecting the electromagnetic field intensity distribution characteristics of the hidden electromagnetic fingerprint RFID tag on the outer packaging of automobile parts in step S3.

[0051] Seven common RFID tag types were used, manufactured by three different manufacturers. Before data collection, the same EPC code was written to all 140 tags. After all tags were written with the same EPC code, data collection was performed. The test material used was authentic and counterfeit automotive brake hoses, which were placed in a tag box for testing.

[0052] The radio frequency signal recognition test equipment used in this experiment is Voyantic Tagformance Pro, such as Figure 11 As shown. The Voyantic Tagformance Pro is a professional radio frequency identification (RFID) test device primarily used for testing ultra-high frequency (UHF) RFID and near-field communication (NFC) high frequency (HF) RFID. Designed for testing tags, cards, and sensor tags, the device supports widely used RFID protocols. It can verify and optimize the sensitivity of UHF and HF tags, study the effects of materials, direction, and distance on tags, and investigate the performance of tags in different applications. The Tagformance Pro is an integrated test device that includes software, accessories, and an anechoic measurement chamber to form a complete test system. Its UHF band supports protocol standards such as ISO 18000-6C (EPC C1G2), ISO 18000-6B, and GB / T 29768; its HF band supports protocol standards such as ISO14443 and ISO 15693.

[0053] The 925MHz standard frequency band was selected for the RFID experiment. Before collecting data from the RFID experiment, a calibration test was conducted to ensure the accuracy of the data. During each data collection, multiple label boxes containing genuine and fake automobile brake hoses were placed in the cardboard box above the antenna device to form a 3×3 multi-label rectangular array, such as Figure 12 The remaining tags were not within the reading system's magnetic field to reduce the risk of tag collisions. All data collection was performed in an isolated environment, which could include thermal noise, cell phone noise, wireless network noise, and RF noise. The RF antenna's transmitted signal frequency was recorded, and sensitivity data was read for each tag.

[0054] The RFID reader is deployed next to the RF antenna, and information from RFID tags within the reading range is collected in real time through an external RFID antenna array. RFID tags are installed on the automotive parts to be sensed, and the data is entered into the neural network. Four antennas are evenly distributed on a gantry placed indoors to scan RFID tags within the reading range from different angles. Each antenna can adjust parameters based on instructions issued by the reader, using different transmission powers for tag scanning. Simultaneously, the energy value of the reflected signal from the RFID tag is accurately collected and used as an identification reference. The RFID antenna array transmits RF signals, and the RFID reader collects the reflected RF signal data in real time through the RFID antenna array. Each set of data consists of the RF signal strength value and the arrival time of the feedback RF signal from the four RFID antennas on the gantry. Time-frequency division multiplexing technology is used to address signal interference when multiple tags in a mixed group of automotive parts respond simultaneously.

[0055] At this frequency, the sensitivity data of each tag of the automobile brake hose are shown in Table 1.

[0056]

[0057] Table 1 records experimental data from RFID tag electromagnetic signature collection for genuine and counterfeit automotive brake hose samples with different EPC codes at the 925MHz standard test frequency band. The table includes the tag name (Name), electronic code (ePC code), sensitivity (in dBm), and backscatter power (in dBm) for both genuine and counterfeit items. Sensitivity reflects the tag's ability to receive RF signals within the frequency band, while backscatter power reflects the strength of the signal reflected by the tag. Comparing the two sets of data for genuine and counterfeit samples reveals that authentic tags generally have higher sensitivity, while backscatter power is generally lower. This difference can be used as a key characteristic parameter for neural network electromagnetic fingerprint recognition, providing a quantitative basis for verifying the authenticity of automotive parts.

[0058] Using the data collected in Table 1, we extracted four types of features and input them into the neural network model. The final analysis results are shown in Table 2:

[0059] As can be seen, the overall accuracy rate reached 98.2%, demonstrating the model's excellent overall discrimination ability and meeting the technical requirements for high-value products. Further analysis of category-specific indicators revealed that the recall rate for genuine products (97.5%) was slightly lower than the recall rate for counterfeit products (98.8%). This difference may be due to the characteristic attenuation of genuine product signals in complex environments (such as metal interference), resulting in some samples being misclassified as counterfeits. Counterfeits, on the other hand, are more easily identified by the model due to their significant electromagnetic characteristics (such as phase shift and spectral entropy change). The high F1-Score (98.0%), the harmonic mean of precision and recall, confirms that the system achieves a balance between reducing false positives and missing detections, highlighting the effectiveness of the dynamic weighting mechanism using the electromagnetic field intensity distribution characteristics of multiple RFID tags combined with a neural network algorithm.

[0060] In order to verify the effectiveness of the algorithm, the algorithm performance is compared with the traditional support vector machine, as shown in Table 3:

[0061] Table 3 shows a comparison of model performance, demonstrating significant advantages in classification tasks: Its accuracy reaches 98.6%, a 14.2 percentage point improvement over the support vector machine (84.4%); its F1 score reaches 0.98, a 7.2% improvement over the support vector machine (SVM). This result demonstrates that combining the electromagnetic field intensity distribution features of multiple RFID tags with a neural network algorithm effectively enhances the model's ability to represent complex features, significantly improving classification accuracy and generalization. Although training time (38.2 seconds) is approximately twice that of the SVM (12.7 seconds), this computational cost is acceptable in accuracy-sensitive scenarios.

[0062] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-source fusion traceability method for automobile parts based on distributed blockchain, characterized in that: The following steps are involved: Step S1: Build blockchain network module: Each business entity in the automotive parts supply chain is added to the blockchain as a node. The business entities include upstream suppliers, manufacturers, downstream distributors, retailers and customers. An authentication center is set up in the blockchain network, with each supplier, manufacturer, and seller acting as an independent authentication center to authenticate the authenticity of product information and the rationality of decisions in the automotive parts supply chain; Construct a multi-source data fusion analysis module as a node to join the blockchain network module. The multi-source data fusion analysis module is used to receive information collection requirements transmitted by various business entities in the supply chain, determine the data set according to the requirements, perform data collection and processing, and then return the processing results to each business entity in the supply chain; Step S2: constructing a multi-source fusion traceability platform module for automobile parts, wherein the multi-source fusion traceability platform module is used to record information of each link in the automobile parts supply chain and use timestamps to record the time sequence; Step S3: Construct a multi-source data acquisition module to form unique identity information of auto parts: Collect microscopic visual feature images of the surfaces of automotive parts and components, and establish a dataset of microscopic visual feature images of the surfaces of various automotive parts and components; Using a multi-antenna array, a radio frequency tag reader collects the electromagnetic field intensity distribution characteristics of the hidden electromagnetic fingerprint RFID tag on the outer packaging of automotive parts, and uses a neural network algorithm to perform feature analysis and classification identification on the extracted electromagnetic fingerprint data; Collect production information, transportation information, inspection information, and import and export information; Step S4, constructing a multi-source data fusion analysis module: sending the data collected in step S3 to the multi-source data fusion analysis module, and using the multi-source data fusion analysis module to process and fuse the data; Step S5: Constructing a data on-chain module: The data merged in step S4 is written to the blockchain through a smart contract module pre-deployed on the blockchain. The smart contract is a computer protocol that automatically executes the terms of the contract and is used to automatically verify and store data. Each transaction record is packaged into a block, which contains the hash value of the current block, the hash value of the previous block, and timestamp information. Step S6: Constructing a traceability query module: The user inputs parts-related information through the multi-source fusion traceability platform module, and the multi-source fusion traceability platform module searches for matching traceability data on the blockchain according to the query rules in the smart contract module.

2. The multi-source fusion traceability method for automobile parts based on distributed blockchain according to claim 1 is characterized by: The blockchain network in step S1 adopts a distributed storage structure, which is used to store different types of basic data. Different types of data are mapped to the block bodies of corresponding blocks in the blockchain. The distributed storage data structure consists of the following five parts: Data layer: refers to supply chain management related data; Classification Merkle Tree Layer: Data of different categories form different Merkle trees, and the Merkle trees do not intersect with each other. This layer is used to store Merkle trees; Classification Merkle tree root layer: This layer stores the roots of different Merkle trees; Directory layer: This layer constructs a directory based on the basic data corresponding to different classification Merkle roots, and calculates a Merkle root from the classification Merkle roots. The number of classification Merkle roots can be selected according to actual needs. Block header layer: All Merkle tree roots in the directory layer are calculated to form the total Merkle tree root. The block header will contain this tree root, including but not limited to the block header hash value of the previous block, the version number of this block, consensus authentication parameters, and timestamp.

3. The multi-source fusion traceability method for automobile parts based on distributed blockchain according to claim 1 is characterized by: The automotive parts multi-source fusion traceability platform module in step S2 adopts a three-layer architecture design, namely a data application layer, a data management layer, and a data collection layer; among them, the data application layer provides traceability information query services for consumers and suppliers, the data management layer is responsible for managing the storage, processing, and maintenance of data required for various business functions, and the data collection layer is responsible for collecting and acquiring relevant data during the automotive parts production process; and in the platform, each business entity only has the authority to distribute the storage blockchain and stores the data in a distributed manner.

4. The multi-source fusion traceability method for automobile parts based on distributed blockchain according to claim 1 is characterized by: In the step S3, the collected microscopic visual feature image of the surface of the automotive parts material is preprocessed, including image segmentation, obtaining a directional map, enhancement and filtering, binarization and refinement; and the texture type is determined according to its two-dimensional frequency domain characteristics, and a corresponding algorithm is selected to extract and analyze key attribute feature points of the texture features; the texture types include continuous, discontinuous, and contour types. For continuous textures, the Poincare index algorithm is selected, for discontinuous textures, the gray-level co-occurrence matrix GLCM algorithm is selected, and for contour textures, the Freeman chain code and its derivative feature extraction algorithm are adopted; and the extracted texture features are analyzed and matched using a deep learning framework CNN.

5. The multi-source fusion traceability method for automobile parts based on distributed blockchain according to claim 1 is characterized by: In step S3, the step of collecting the electromagnetic field intensity distribution characteristics of the hidden electromagnetic fingerprint RFID tag includes: Step S3-1: Perform IQ demodulation on the RFID tag signal received by the RFID tag reader to obtain I and Q signals, and obtain the amplitude signal after modulus calculation. ,in, , is the sampling point; Step S3-2: Perform clustering to obtain the center point vector corresponding to code element 0 and 1 ,in is the cluster center corresponding to code element 0, is the cluster center point corresponding to code element 1, and the clustering function is expressed as: ,in, is the clustering operation function; Step S3-3: Determine based on clustering of silent period signal and EPC signal and , the expected signal is obtained by Euclidean distance judgment ,in, , is the Euclidean distance decision function; Step S3-4: Amplitude signal Normalize to remove amplitude differences: ， in, is a standard amplitude signal; Step S3-5: Standard amplitude signal Removing the desired signal To calculate the noise signal : ， in, is a noise signal; Step S3-6: Expected signal , standard amplitude signal , noise signal and the original signal By calculating the cross spectral entropy and related features through pairwise combination, the cross spectral entropy features are extracted, and the spectral entropy Defined as: ， in, For signal In frequency The energy proportion probability density function at : ， The cross power spectral density is expressed based on the Wiener-Schinchin theorem as: ， in, For signal and The cross power spectrum reflects the correlation between signals in the frequency domain. for The conjugate of represents the Fourier transform, For signal and The cross-correlation function is used to evaluate the cross-correlation between two signals. The similarity of is the delay variable; From the above formula, the cross spectral entropy can be obtained as: ， in, For signal and The cross spectral entropy of is used as the feature of RFID tags for automotive parts; is the energy proportion probability density function of the cross power spectrum; At the same time, the introduction of the maximum , maximum phase , maximum frequency , and cross spectral entropy Together they serve as the characteristics of RFID tags for automotive parts; their expressions are: 。 6. The multi-source fusion traceability method for automobile parts based on distributed blockchain according to claim 5 is characterized by: The multi-source fusion features of the parts calculated in step S3-6 are processed using a neural network algorithm, specifically including: , maximum phase , maximum frequency , and cross spectral entropy As the input of the neural network, the network is trained by introducing a back-propagation algorithm optimized by the momentum term. After the training is completed, the network is used to perform classification inference on the input fusion features, output the authenticity determination results of the parts and the corresponding traceability information, and write the results into the distributed blockchain to complete the evidence storage.

7. The multi-source fusion traceability method for automobile parts based on distributed blockchain according to claim 1 is characterized by: In step S5, the smart contract module completes decision-making activities and performs distributed accounting through decentralized consensus authentication, making the smart contract and product data traceable; in the storage link, the user calls the smart contract to trace the product information on the chain; the uploaded information will be recorded in the blockchain ledger and a transaction hash address will be returned as the unique identifier of the information in the blockchain.

8. The multi-source fusion traceability method for automobile parts based on distributed blockchain according to claim 1 is characterized by: Each business entity only has the authority to store distributed blockchain data. They can only perform data storage operations after encrypting their identity data with encryption keys. They do not have the right to store business-related data that has a mapping relationship with the blockchain. The encryption keys are collaboratively managed and backed up by multiple authoritative automotive parts industry organizations. When verifying the authenticity of automotive parts, the encryption key provided by the supply chain entity and the official backup key are used to decrypt the relevant product data respectively, and the consistency of the two decryption results is compared to determine whether the supply chain entity has tampered with the product data key.

9. The multi-source fusion traceability method for automobile parts based on distributed blockchain according to claim 1 is characterized by: The automotive parts supply chain is built on an industry-specific network or the Internet, and only business entities in the supply chain are authorized to join the blockchain system.

10. A distributed blockchain-based multi-source fusion traceability system for automobile parts, applied to the distributed blockchain-based multi-source fusion traceability method for automobile parts according to claims 1 to 9, characterized in that: include: Blockchain network module: used to build a blockchain network, adding certification centers, suppliers, manufacturers, sellers, and multi-source data fusion analysis centers as nodes to achieve information sharing and supervision; Auto parts multi-source fusion traceability platform module: used to build an auto parts multi-source fusion traceability platform; Multi-source data acquisition module: used to collect component information at all stages of the automotive parts process, including production time, batch, raw material source, processing technology, transportation route, and sales records; Multi-source data fusion analysis module: used to process and fuse the collected data to form complete traceability information; Data on-chain module: used to write the integrated data into the blockchain through smart contracts to ensure that the data cannot be tampered with and is traceable; Traceability query module: used to receive user query requests, search for matching traceability data on the blockchain according to the query rules in the smart contract, and display it to the user; Smart contract module: used to achieve traceability and automatic collection of automotive parts supply chain information, and complete decision-making activities and conduct distributed accounting through decentralized consensus authentication based on the automatic operation of the blockchain.

Images