Cleaning and spraying data management system based on block chain

Abstract

The invention relates to the technical field of blockchain data management, and discloses a blockchain-based cleaning and spraying data management system, which comprises a cleaning parameter acquisition module, a data fragment encryption module, a cross-chain feature verification module, a dynamic decision reconstruction module and an execution instruction generation module. The system obtains the operation state data flow of the spraying equipment through distributed nodes, generates a data verification feature matrix set through space-time dimension division and feature fragmentation encryption, obtains an optimized verification feature matrix set through consensus mechanism screening and feature relevance verification, generates a data chain type decision feature map based on Hash time lock fusion, and obtains a data chain type decision feature map. And finally, an equipment control instruction set is generated. The system realizes data security storage, cross-chain verification and dynamic decision by means of a block chain technology, improves the cleaning and spraying data management efficiency and decision accuracy, is suitable for intelligent data management and equipment control in an industrial scene, and solves the problems of weak data security, decision lag and the like of a traditional system.

Description

Technical Field

[0001] The present invention relates to the technical field of blockchain data management, and specifically to a cleaning and spraying data management system based on blockchain. Background Art

[0002] In the fields of industrial cleaning and spraying, the efficient management and reliable control of equipment operation data are the core links to ensure production quality and improve operation efficiency. Traditional data management systems generally have problems such as weak data security protection, insufficient analysis of the correlation of multi-source data, and lagging decision-making responses, making it difficult to meet the requirements of modern intelligent manufacturing for data real-time, security, and intelligent decision-making.

[0003] From the perspective of data security, traditional systems usually adopt a centralized storage architecture, and data is vulnerable to threats such as single-point failures, network attacks, or human tampering. For example, during the operation of spraying equipment, if key parameters such as pressure, temperature, and spraying flow rate are illegally tampered with, it may lead to quality problems such as uneven coating thickness of products and incomplete cleaning, and even cause equipment failures or safety accidents. In addition, traditional encryption methods mostly target single data dimensions and lack a hierarchical encryption mechanism for spatio-temporal correlated data, making it difficult to cope with complex data leakage risks.

[0004] In terms of data processing and analysis, traditional systems often lack the ability to effectively integrate and deeply mine multi-source heterogeneous data. Cleaning and spraying equipment generates data streams with multiple dimensions such as time series, spatial location, and equipment status during operation. Traditional methods usually only perform simple time series analysis on the data, ignoring the correlation of data in the spatio-temporal dimension. For example, there may be potential correlations between the fluctuations of equipment operation parameters at different time periods and the changes in workshop environmental temperature and humidity, but traditional systems cannot efficiently capture such correlation features, resulting in limited accuracy and reliability of the decision-making model.

[0005] From the perspective of decision-making and control mechanisms, traditional systems mostly use preset rules or static models for decision-making and lack the ability to respond in real time to dynamic changes. When the equipment operation state or production environment changes suddenly, the static decision-making model is difficult to quickly adjust the control strategy, which may lead to a decrease in production efficiency or waste of resources. For example, during the spraying process, if there are slight changes in the workpiece material, traditional systems cannot timely adjust the spraying parameters according to real-time data, which may cause waste of paint or unqualified coating quality.

[0006] In addition, traditional systems have significant deficiencies in cross-device and cross-system data interaction. With the development of the industrial Internet, cleaning and spraying equipment often needs to share data and collaborate with systems in other production links. However, the closed architecture of traditional systems leads to poor data circulation, making it difficult to achieve cross-chain data verification and collaborative decision-making. For example, in the scenario of collaborative operation of multiple production lines, the data of each device cannot be effectively interconnected and verified, which may lead to disconnection of the production process or difficulties in quality traceability. Summary of the Invention

[0007] The purpose of the present invention is to provide a cleaning and spraying data management system based on blockchain to solve the problems proposed in the above background technology.

[0008] To achieve the above purpose, the present invention provides the following technical solution: A cleaning and spraying data management system based on blockchain, the system includes: A cleaning parameter acquisition module, used to obtain the operation status data stream of the spraying equipment through distributed nodes; A data sharding and encryption module, used to perform spatio-temporal dimension division and feature sharding encryption on the operation status data stream to generate a set of data verification feature matrices; A cross-chain feature verification module, used to perform consensus mechanism screening and feature correlation verification on the set of data verification feature matrices to obtain a set of optimized verification feature matrices; A dynamic decision reconstruction module, used to perform feature fusion based on hash time lock on the set of optimized verification feature matrices to generate a data chain decision feature map; An execution instruction generation module, used to generate a device control instruction set according to the data chain decision feature map.

[0009] Preferably, the data sharding and encryption module includes: A data stream splitting unit, used to dynamically split the operation status data stream according to a preset block size to form a set of data shards; A feature encryption unit, used to input each shard in the set of data shards into a data sharding encryptor based on a Merkle tree to generate the set of data verification feature matrices.

[0010] Preferably, the cross-chain feature verification module includes: A feature matrix deconstruction unit, used to perform chain parsing on the data verification feature matrix along the time stamp dimension to obtain a set of feature verification vectors; A consensus association calculation unit, used to calculate the consensus verification value between any two feature vectors in the set of feature verification vectors to generate a feature association consensus matrix; A dynamic screening unit for adjusting the weights of the feature correlation consensus matrix according to the hash differences between adjacent blocks in the set of feature verification vectors to generate a consensus constraint correlation matrix; A feature optimization unit for performing cross-chain convolution aggregation on the consensus constraint correlation matrix and the set of feature verification vectors to obtain the optimized verification feature matrix.

[0011] Preferably, the consensus correlation calculation unit includes: A feature mapping sub-unit for mapping each feature vector in the set of feature verification vectors to an orthogonal hash space to obtain a set of mapped feature vectors; A consensus correlation analysis sub-unit for calculating the hash similarity between any two mapped feature vectors in the set of mapped feature vectors to generate the feature correlation consensus matrix composed of multiple consensus correlation values.

[0012] Preferably, the feature optimization unit is specifically implemented as: Performing chained convolution processing on the consensus constraint correlation matrix to generate a consensus constraint feature tensor; Inputting the set of feature verification vectors and the consensus constraint feature tensor into a bidirectional timestamp encoding network to generate a set of chained context feature vectors; Performing matrix recombination on the set of chained context feature vectors to generate the optimized verification feature matrix.

[0013] Preferably, the dynamic decision reconstruction module includes: A feature matrix compression unit for performing max pooling processing on each matrix in the set of optimized verification feature matrices in the timestamp dimension to generate a set of optimized verification feature vectors; A hash entropy value calculation unit for calculating the hash entropy value of each feature vector in the set of optimized verification feature vectors to generate a dynamic hash entropy set; A benchmark feature determination unit for selecting the optimized verification feature vector corresponding to the maximum hash entropy value in the dynamic hash entropy set as the initial decision benchmark vector; A dynamic weight calculation unit for calculating the dynamic decision weights of each feature vector according to the hash distance between each feature vector in the set of optimized verification feature vectors and the initial decision benchmark vector and the hash entropy value of each feature vector to generate a dynamic weight set; A chained reconstruction unit for performing weighted aggregation on the set of optimized verification feature vectors by using the dynamic weight set to generate the data chained decision feature map.

[0014] Preferably, the hash entropy value calculation unit is specifically implemented as: Calculate the mean vector and variance vector of the optimized verification feature vector; Perform element-wise difference calculation between the optimized verification feature vector and the mean vector, and perform a square operation on the difference result to generate a feature hash difference vector; Calculate the overall median of the feature hash difference vector; perform a ratio operation on the median and the squared value of the variance vector, and input the result into a normalization function to obtain the hash entropy value.

[0015] Preferably, the dynamic weight calculation unit is specifically implemented as: Multiply the hash entropy value of the optimized verification feature vector by a first decision coefficient with the hash entropy value of the initial decision benchmark vector to obtain a first dynamic decision factor; Multiply the absolute value of the Euclidean distance between the optimized verification feature vector and the initial decision benchmark vector by a second decision coefficient to obtain a second dynamic decision factor; Perform weighted summation on the first dynamic decision factor and the second dynamic decision factor to obtain the dynamic decision weight.

[0016] Preferably, the execution instruction generation module is specifically implemented as: Input the data chain decision feature map into an instruction generator based on a smart contract to generate the device control instruction set, and the instruction set is used to control the operating parameters of the cleaning and spraying equipment.

[0017] Preferably, the spatio-temporal dimension division includes: Slice the operation status data stream into continuous data segments according to timestamps, and attach spatial grid coding to each segment to form a spatio-temporal associated shard set; The feature shard encryption performs cross-signature verification on the shard set through the asymmetric keys of distributed nodes.

[0018] Compared with the prior art, the beneficial effects of the present invention are: In terms of data security protection, the system realizes the spatio-temporal dimension division and feature shard encryption of the operation status data stream through the data shard encryption module. The data stream is dynamically divided into a data shard set according to a preset block size, and each shard is encrypted using a Merkle tree structure. Combined with the asymmetric key cross-signature verification mechanism of distributed nodes, a multi-level data security protection system is formed. This encryption method not only realizes fine-grained protection of data, but also, through the distributed storage feature of the blockchain, disperses the data shards and stores them on different nodes, significantly reducing the risk of single-point leakage and ensuring the integrity and immutability of data during the processes of data collection, transmission, and storage. For example, when a data shard of a certain node is attacked, other nodes can still verify the authenticity of the data through the consensus mechanism, ensuring the stable operation of the system.

[0019] In the data processing and verification phase, the cross-chain feature verification module realizes the in-depth mining and optimization of multi-dimensional data features through the chained parsing, consensus association calculation, dynamic screening, and cross-chain convolution aggregation of the data verification feature matrix. By parsing the feature verification vector set along the timestamp dimension, calculating the hash similarity between any two feature vectors to generate a feature association consensus matrix, and adjusting the weights according to the hash differences between adjacent blocks, the module effectively captures the correlation and dynamic change law of data in the spatio-temporal dimension. By generating chained context feature vectors through a two-way timestamp encoding network, the module further enhances the context semantic understanding ability of data features, enabling the system to extract a more representative optimized verification feature matrix from complex data streams and providing more accurate data support for subsequent decision-making. For example, when analyzing the fluctuations in the pressure parameters of a spraying device, the system can discover potential correlations between them and parameters such as paint viscosity and environmental temperature through the feature association consensus matrix, thus more comprehensively evaluating the operating state of the device.

[0020] The dynamic decision-making reconstruction module realizes a dynamic decision-making mechanism based on real-time data through operations such as max pooling, hash entropy value calculation, dynamic weight assignment, and weighted aggregation. Selecting the feature vector corresponding to the maximum hash entropy value as the initial decision-making benchmark, and calculating dynamic weights by combining the hash distances between each feature vector and the benchmark vector and their own hash entropy values, the module enables the system to automatically adjust decision-making weights according to the uncertainty and correlation of data, generating a data chained decision-making feature map that better fits the actual operating state. This dynamic decision-making mechanism significantly improves the system's adaptability to complex working conditions and can quickly respond to changes in the device operating state or production environment. For example, when the workpiece material changes and the spraying parameters need to be adjusted, the system can re-aggregate the feature vectors according to the dynamically calculated weights in real time, quickly generating a new decision-making feature map to guide the device to adjust the operating parameters and avoid quality problems or resource waste caused by static decision-making.

[0021] The execution instruction generation module converts the decision-making feature map into a device control instruction set through a smart contract, realizing the automatic connection between decision-making and control. The programmability and automatic execution characteristics of the smart contract ensure the accuracy and timeliness of instruction generation. At the same time, the consensus mechanism of the blockchain guarantees the immutability and traceability of the instructions, ensuring that the control instructions can be transmitted to the cleaning and spraying device safely and reliably to achieve precise regulation of the operating parameters. For example, during the cleaning process, the system can generate instructions to adjust parameters such as the flow rate and pressure of the cleaning liquid according to real-time data, ensuring the stability and consistency of the cleaning effect.

[0022] In addition, through cross-chain feature verification and the collaborative work of distributed nodes, the system realizes data interconnection and verification across devices and systems, breaks the closed architecture of traditional systems, and provides strong support for multi-device collaborative operations and quality traceability in the industrial Internet environment. Through the chain structure and timestamp mechanism of the blockchain, the system can completely record the generation, processing, and decision-making processes of data, providing a reliable basis for the transparent management of the production process and the rapid traceability of quality problems. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a working schematic diagram of the blockchain-based cleaning and spraying data management system described in the present invention; Figure 2 It is a design diagram of the data sharding and encryption module; Figure 3 It is a design diagram of the cross-chain feature verification module; Figure 4 It is a design diagram of the dynamic decision-making reconstruction module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0024] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0025] Please refer to Figures 1-4 , the blockchain-based cleaning and spraying data management system involved in the present invention, the system includes: The present invention provides a blockchain-based cleaning and spraying data management system. The system includes a cleaning parameter acquisition module, a data sharding and encryption module, a cross-chain feature verification module, a dynamic decision-making reconstruction module, and an execution instruction generation module. Each module collaborates to achieve blockchain management of cleaning and spraying data. The specific implementation is as follows: Cleaning parameter acquisition module: Deployed in the spraying equipment network through distributed nodes, it can obtain the operation status data stream of the spraying equipment in real time, including but not limited to multi-dimensional data such as equipment operation temperature, pressure, spraying flow rate, operation timestamp, and equipment geographical location coordinates. Based on the P2P network architecture of the blockchain, the distributed nodes ensure the decentralization and real-time nature of data stream acquisition. Each node has the functions of data acquisition and preliminary verification to prevent data loss caused by single-point failures.

[0026] Data Sharding Encryption Module: It performs spatio-temporal dimension division and feature sharding encryption on the collected operation status data stream. The spatio-temporal dimension division is specifically to slice the data stream into continuous data segments according to timestamps, and each segment of data is appended with a spatial grid code (such as a grid index based on the device's GPS coordinates) to form a set of spatio-temporally associated shards; Feature sharding encryption cross-signs and verifies the shard set through the asymmetric keys of distributed nodes, and uses the Merkle tree structure to generate a set of data verification feature matrices to achieve encrypted storage and fast verification of data shards.

[0027] Cross-chain Feature Verification Module: It performs consensus mechanism screening and feature correlation verification on the set of data verification feature matrices. First, it effectively screens the feature matrices submitted by each node through a consensus mechanism (such as an improved PBFT algorithm) and eliminates the matrices corresponding to abnormal data; Then, based on the hash similarity calculation and cross-chain convolution aggregation of feature vectors, it verifies the correlation between features and generates a set of optimized verification feature matrices to ensure the consistency and reliability of data features.

[0028] Dynamic Decision Reconstruction Module: It performs feature fusion based on hash time locks on the set of optimized verification feature matrices. It compresses the matrix dimension through max pooling processing in the timestamp dimension to generate a set of feature vectors; Based on the calculation of hash entropy values, it determines the initial decision benchmark vector, dynamically calculates the weights of each feature vector in combination with hash distance and entropy values, and finally generates a data chain decision feature map through weighted aggregation to achieve dynamic reconstruction and decision correlation of data features.

[0029] Execution Instruction Generation Module: It inputs the data chain decision feature map into an instruction generator based on a smart contract, and generates a device control instruction set through preset control logics (such as PID control algorithm parameters, device start-stop rules, etc.). The instruction set is directly used to control the operation parameters of the cleaning and spraying equipment, such as spraying pressure adjustment, cleaning liquid flow control, operation path planning, etc., to achieve data-driven device automation control.

[0030] The present invention will be further described below in conjunction with Embodiments 1 to 5: Embodiment 1: This embodiment relates to the specific implementation of a data sharding encryption module and a cross-chain feature verification module. The data sharding encryption module includes a data stream splitting unit and a feature encryption unit. The data stream splitting unit needs to dynamically split the running state data stream. Specifically, according to a preset block size (which can be adaptively adjusted according to actual situations such as network bandwidth and data traffic, for example, initially set to 512KB), the continuous data stream is divided into multiple data shards, forming a data shard set. Each data shard needs to carry information such as a timestamp, a spatial grid code, and a shard sequence number. Among them, the timestamp is used to identify the time point of data collection, the spatial grid code can be generated based on the GPS coordinates of the device through the grid division algorithm of the geographic information system (for example, the area where the device is located is divided into grid cells with a specific resolution, such as 50m×50m, and each grid cell corresponds to a unique code), and the shard sequence number is used to reflect the order of the data shard in the original data stream. Through these information, the spatio-temporal correlation of each data shard is ensured, facilitating subsequent data processing and verification.

[0031] The role of the feature encryption unit is to process the data shard set by inputting it into a data shard encryptor based on the Merkle tree. The specific process is as follows: using a hash algorithm (such as the SHA-256 hash algorithm) to generate a corresponding hash value for each data shard, and then according to the construction rules of the Merkle tree, these hash values are combined and calculated layer by layer, finally constructing a Merkle tree. In this process, the hash value of each data shard will form its hash path in the Merkle tree, and these hash path information will be integrated to generate a set of data verification feature matrices. The root hash value of the Merkle tree will be stored together with the data shards in each node of the blockchain. When it is necessary to verify the data integrity, only by comparing the root hash value can it be quickly determined whether the data has been tampered with. This method can achieve the encrypted storage and quick verification of data shards, ensuring the security and reliability of data during storage and transmission.

[0032] The cross-chain feature verification module includes a feature matrix deconstruction unit, a consensus association calculation unit, a dynamic screening unit, and a feature optimization unit. The feature matrix deconstruction unit needs to perform chain parsing on the data verification feature matrix along the timestamp dimension. Specifically, each data verification feature matrix is disassembled into a series of feature verification vectors in chronological order. These vectors are arranged in chronological order to form a set of feature verification vectors. Each vector corresponds to the device running characteristics under a time slice. For example, within a 100ms time window, the characteristic data such as the running temperature, pressure, and flow of the device will be integrated into a feature verification vector. In this way, the matrix-form data is converted into a vector sequence that is convenient for analysis, laying a foundation for subsequent feature correlation verification.

[0033] The processing flow of the consensus association calculation unit is first to map each feature vector in the set of feature verification vectors to an orthogonal hash space. The mapping of the orthogonal hash space can be achieved through specific mathematical transformations. For example, a set of linearly independent hash functions are used to project the feature vectors, converting them into frequency-domain feature vectors or other forms of feature vectors, thereby reducing the linear correlation between feature vectors and making the features more distinguishable and analyzable in the hash space. After the mapping is completed, it is necessary to calculate the hash similarity between any two mapped feature vectors in the set of mapped feature vectors. The calculation of the hash similarity can adopt various algorithms, such as the cosine similarity algorithm, the Hamming distance algorithm, etc. The hash similarity values obtained through calculation will generate a feature association consensus matrix. The rows and columns of this matrix respectively correspond to the indices of the feature verification vectors, and the element values in the matrix reflect the association strength between two feature vectors. The larger the value, the stronger the association between the features, and vice versa. Through this matrix, the mutual relationship between features can be intuitively understood.

[0034] The function of the dynamic screening unit is to adjust the weights of the feature association consensus matrix according to the hash differences between adjacent blocks in the set of feature verification vectors. The hash differences between adjacent blocks can be obtained by calculating the forward difference, that is, calculating the change rate of the hash values of two adjacent blocks. For features with significant hash differences, it indicates that their changes in the time series are relatively large and they may be key features. Therefore, it is necessary to increase their weights in the feature association consensus matrix; while for features with small hash differences, their weights are correspondingly reduced. Through this way of weight adjustment, a consensus constraint association matrix is generated. This matrix can highlight the influence of key features and make the subsequent feature optimization process pay more attention to the features that play an important role in decision-making.

[0035] The implementation process of the feature optimization unit is relatively complex. First, perform chained convolution processing on the consensus constraint correlation matrix. Chained convolution processing can be carried out in the form of three-dimensional convolution (the size of the convolution kernel can be set according to actual needs, such as 3×3×3). Through the convolution operation, extract the feature correlation patterns in the spatial and temporal dimensions to generate the consensus constraint feature tensor. Then, input the set of feature verification vectors and the consensus constraint feature tensor into a bidirectional timestamp encoding network, which can be a network structure with the ability to process time series data such as a Bi-LSTM network. The input layer of the network receives the feature vectors and timestamp encoding information. The hidden layer processes the time series through bidirectional recurrent neurons, which can capture the feature dependency relationships of past and future time steps, so as to fully mine the context information in the data. The output layer will generate a set of chained context feature vectors containing context information. Finally, perform matrix recombination operations on the set of chained context feature vectors, that is, arrange the multi-dimensional vectors in chronological order into a two-dimensional matrix. In this way, the set of chained context feature vectors is converted into an optimized verification feature matrix. The rows of this matrix correspond to the time series, and the columns correspond to the feature dimensions. Through this processing, the spatio-temporal correlation optimization and cross-chain aggregation of features are realized, so that the optimized feature matrix can more accurately reflect the change law of the device operation state, providing a more reliable basis for subsequent dynamic decision-making reconstruction.

[0036] During the processing of the entire data sharding encryption module and the cross-chain feature verification module, each unit cooperates closely. The data flow splitting unit provides a structured set of data shards for the feature encryption unit, and the feature encryption unit realizes the encryption and verification of data through the Merkle tree structure; each unit in the cross-chain feature verification module then parses, maps, calculates similarity, adjusts weights, and optimizes features for the data verification feature matrix in turn, gradually screening and optimizing features to ensure that the finally obtained set of optimized verification feature matrices can accurately and reliably reflect the feature correlation and spatio-temporal characteristics of the device operation state data, laying a solid foundation for the subsequent processing links of the entire blockchain-based cleaning and spraying data management system. Through such a design, the system can efficiently process and verify data in a distributed environment, ensuring the integrity, security, and reliability of the data, so as to realize the effective management and utilization of the operation data of the cleaning and spraying equipment.

[0037] Embodiment 2: This embodiment relates to a further refined implementation of a consensus association calculation unit and a feature optimization unit. The consensus association calculation unit includes a feature mapping subunit and a consensus correlation analysis subunit. The core function of the feature mapping subunit is to map each feature vector in the set of feature verification vectors to an orthogonal hash space to reduce the linear correlation between features, facilitating subsequent similarity calculation and feature analysis. In specific implementation, a set of predefined orthogonal hash function groups can be used. These hash functions have the property of linear independence and can project the original feature vectors from the original feature space to a multi-dimensional orthogonal hash space. For example, for the original feature verification vectors containing multi-dimensional features such as temperature, pressure, and flow rate, through the action of the orthogonal hash function group, they can be converted into frequency-domain feature vectors or other forms of orthogonal feature vectors, so that in the new hash space, the correlation between different feature dimensions is significantly reduced, thus more clearly showing the independent characteristics and mutual relationships of each feature.

[0038] The consensus correlation analysis subunit calculates the hash similarity between any two mapped feature vectors based on the mapped feature vectors, and then generates a feature association consensus matrix. Multiple algorithms can be used to calculate the hash similarity, such as the Hamming distance algorithm or the cosine similarity algorithm. Taking the cosine similarity algorithm as an example, its basic principle is to measure the similarity between two vectors by calculating the cosine value of the angle between them. The closer the cosine value is to 1, the more similar the two vectors are, that is, the stronger the correlation between features; conversely, the closer the cosine value is to -1 or 0, the weaker the correlation or the lack of correlation. In practical applications, first, the mapped feature vectors are normalized to eliminate the influence of dimensional differences, and then the cosine similarity values between any two vectors are calculated pairwise. These values are arranged in the index order of the feature vectors to form a two-dimensional matrix, that is, the feature association consensus matrix. The rows and columns of this matrix correspond to different feature verification vectors respectively, and each element value in the matrix is a numerical value between 0 and 1, intuitively reflecting the association strength between any two feature vectors and providing a quantitative basis for subsequent feature screening and optimization.

[0039] The processing flow of the feature optimization unit includes chained convolution processing of the consensus constraint association matrix, feature extraction of the bidirectional timestamp encoding network, and matrix recombination operations. First, chained convolution processing is performed on the consensus constraint association matrix. This processing process can adopt three-dimensional convolution. By setting a convolution kernel of a specific size (such as 3×3×3), sliding convolution operations are performed on the matrix in the spatial and temporal dimensions to extract the feature association patterns therein. Chained convolution processing can capture the local association information between adjacent feature vectors in the matrix and the dynamic change patterns of features in different time slices, generating a consensus constraint feature tensor. This tensor not only contains the feature association information in the original matrix but also enhances the spatial and temporal correlation between features through convolution operations, providing richer context information for subsequent feature fusion.

[0040] Next, the set of feature verification vectors and the consensus constraint feature tensor are input into the bidirectional timestamp encoding network. This network typically adopts a variant structure of the recurrent neural network (RNN), such as the Bi-LSTM network. Its core advantage lies in being able to capture the feature dependency relationships of past and future time steps in time series data simultaneously. The input layer of the network receives two parts of input: one is the feature verification vector after mapping processing, and the other is the corresponding timestamp encoding information. The timestamp encoding can adopt one-hot encoding or other appropriate encoding methods to identify the time order of the feature vectors. The hidden layer consists of bidirectional recurrent neurons. The forward neurons process the time series from front to back to capture the feature information of past time steps; the backward neurons process the time series from back to front to capture the feature information of future time steps. The outputs of both are fused in the hidden layer to generate an intermediate feature representation containing context information. The output layer then generates a set of chained context feature vectors based on the output of the hidden layer. Each vector integrates the feature dependency relationships before and after this time point and can more comprehensively reflect the time series characteristics of the device operating state.

[0041] Finally, a matrix reorganization operation is performed on the set of chained context feature vectors. The specific way of matrix reorganization is to arrange the multi-dimensional chained context feature vectors in chronological order and convert them into a two-dimensional optimized verification feature matrix. For example, if each chained context feature vector is an n-dimensional vector and there are m time step vectors in total, the reorganized matrix is m rows and n columns, where each row corresponds to the feature vector of a time step and each column corresponds to a feature dimension. Through this matrix reorganization operation, the scattered feature vectors are integrated into a structured matrix form, which is convenient for the subsequent dynamic decision-making reconstruction module to perform feature fusion and decision analysis. The optimized verification feature matrix not only retains the spatio-temporal information of the original features but also enhances the correlation and context dependence between features through chained convolution processing and feature extraction by the bidirectional timestamp encoding network, enabling each element in the matrix to more accurately reflect the comprehensive information of the device operating state at a specific time and feature dimension.

[0042] During the collaborative work process of the consensus association calculation unit and the feature optimization unit, the feature mapping subunit provides a more efficient feature representation for feature similarity calculation through the transformation of the orthogonal hash space. The consensus correlation analysis subunit provides a clear basis for feature screening by quantifying the association strength between features. The feature optimization unit realizes the spatio-temporal association optimization and cross-chain aggregation of features through deep feature extraction of chain convolution processing and bidirectional timestamp encoding network. This layer-by-layer processing method, from feature space transformation to similarity calculation and then to deep feature fusion, forms a complete feature verification and optimization process, ensuring that the cross-chain feature verification module can output a high-quality set of optimized verification feature matrices, providing reliable feature data support for the dynamic decision-making and equipment control of the entire blockchain-based cleaning and spraying data management system. Through such a design, the system can effectively mine and utilize the deep feature associations in the equipment operation data in a complex distributed environment, improve the accuracy of data processing and the scientific nature of decision-making, thereby realizing the intelligent and precise management of cleaning and spraying equipment.

[0043] Embodiment 3: This embodiment relates to the specific implementation of the dynamic decision-making reconstruction module, which includes a feature matrix compression unit, a hash entropy value calculation unit, a reference feature determination unit, a dynamic weight calculation unit, and a chain reconstruction unit. The feature matrix compression unit needs to perform maximum pooling processing on each matrix in the set of optimized verification feature matrices in the time dimension. Specifically, for each optimized verification feature matrix, its rows correspond to the time series and its columns correspond to the feature dimensions. The maximum pooling processing is to take the maximum value of all time point data in each column for each feature dimension (column) in the time dimension (i.e., the row direction), thereby compressing the two-dimensional matrix into a one-dimensional optimized verification feature vector, forming a set of optimized verification feature vectors. This processing method can retain the key peak information of each feature dimension in the time series, eliminate redundant time details, realize the compression of matrix dimensions and the extraction of key features, facilitating subsequent entropy value calculation and weight assignment.

[0044] The hash entropy value calculation unit is used to calculate the hash entropy value of the optimized verification feature vector, and its processing process is as follows: First, calculate the mean vector and variance vector of the optimized verification feature vector. Each element of the mean vector is the average value of the corresponding feature dimension, reflecting the average level of this feature; each element of the variance vector is the variance of the corresponding feature dimension, reflecting the fluctuation degree of this feature. Then, perform element-by-element difference calculation on the optimized verification feature vector and the mean vector to obtain the difference vector , and then perform a square operation on the difference result to generate the feature hash difference vector , each element of the vector represents the square of the deviation degree of the corresponding eigenvalue from the mean value, and the larger the value, the farther the eigenvalue deviates from the mean value. Subsequently, calculate the overall median of the feature hash difference vector , as a robust statistic, the median can reduce the influence of extreme values on the result and represent the central tendency of the feature distribution. Finally, calculate the ratio of the median to the square value of the variance vector , after obtaining the intermediate result, input it into a normalization function (such as the Sigmoid function), and map the result to interval to obtain the hash entropy value . The calculation formula of the hash entropy value can be expressed as: ; where, represents the normalization function, which is used to convert the ratio result into a standardized entropy value. The larger the entropy value, the higher the uncertainty of the feature, and higher weights may need to be assigned during the decision-making process.

[0045] The role of the benchmark feature determination unit is to select the optimized verification feature vector corresponding to the maximum hash entropy value from the dynamic hash entropy set as the initial decision benchmark vector . This vector corresponds to the feature state with the highest uncertainty in the data. Using it as a decision reference point can ensure that the decision-making process fully considers the most variable and uncertain features in the data, and avoid decision-making biases caused by ignoring key changing features. For example, if the hash entropy value of a certain feature vector is the largest, it means that the corresponding device operating state fluctuates greatly or is more dispersed in the time series, which may reflect abnormal operating conditions or key operation stages during device operation. Setting it as the benchmark vector helps to highlight the influence of such features during subsequent weight calculation.

[0046] The dynamic weight calculation unit needs to calculate the dynamic decision weights of each feature vector according to the hash distance between the optimized verification feature vector and the initial decision benchmark vector and the hash entropy values of each feature vector. The specific steps are as follows: First, multiply the hash entropy value of the optimized verification feature vector by the hash entropy value of the initial decision benchmark vector and the first decision coefficient (initially set to 0.6, which can be adjusted according to actual needs) to obtain the first dynamic decision factor . This factor reflects the influence of the difference in entropy values between the feature vector and the benchmark vector on the weight. If , then is positive, increasing the weight; otherwise, decreasing the weight. Secondly, calculate the absolute value of the Euclidean distance between the optimized verification feature vector and the initial decision benchmark vector, and multiply it by the second decision coefficient (If the initial setting is 0.4, it can be dynamically adjusted) to obtain the second dynamic decision factor , which reflects the influence of the distance between the feature vector and the reference vector in the feature space on the weight. The greater the distance, the greater the feature difference, and the possible different contributions to the decision. Finally, the first dynamic decision factor and the second dynamic decision factor are weighted and summed to obtain the dynamic decision weight . The calculation formula is as follows: ; Among them, and are weight coefficients used to adjust the influence degrees of entropy difference and spatial distance on the weight. The sum of the two is usually 1 (such as ) to ensure the balance of weight calculation. In this way, the dynamic decision weight of each feature vector takes into account both its uncertainty (entropy value) and the difference from the reference vector (spatial distance), making the weight assignment more scientific and reasonable.

[0047] The chain reconstruction unit uses the dynamic weight set to perform weighted aggregation on the set of optimized verification feature vectors to generate a data chain decision feature map. When specifically implemented, each optimized verification feature vector is multiplied by its corresponding dynamic decision weight to obtain the weighted feature vector . Then, all the weighted feature vectors are summed to finally obtain the data chain decision feature map . This feature map is presented in matrix form. Its rows can correspond to different decision factors or feature combinations, and its columns correspond to feature dimensions. Each element value in the matrix reflects the comprehensive decision influence of the corresponding feature after weighted aggregation. For example, if a certain feature dimension has a high hash entropy value and a distance from the reference vector in the feature vectors at multiple time steps, its corresponding element value in the feature map will increase significantly due to weighted aggregation, indicating that this feature has a high contribution to the decision and needs to be focused on in subsequent device control.

[0048] In the entire processing flow of the dynamic decision-making reconstruction module, the feature matrix compression unit quickly extracts key features through max pooling. The hash entropy value calculation unit quantifies the uncertainty of features through statistics and normalization processing. The benchmark feature determination unit uses high-entropy features as decision anchor points. The dynamic weight calculation unit scientifically allocates weights by combining entropy value differences and spatial distances. The chained reconstruction unit generates a comprehensive decision feature map through weighted aggregation. Each unit is closely linked, from feature compression, entropy value calculation to weight allocation and feature aggregation, forming a complete dynamic decision chain, ensuring that the system can dynamically adjust decision-making basis according to the real-time features of device operation data, generate a decision feature map more in line with the actual working conditions, provide accurate input data for the execution instruction generation module, and then achieve precise control of the operation parameters of the cleaning and spraying equipment. This design not only makes full use of the advantages of distributed storage and verification of blockchain data, but also improves the adaptability of the system to complex working conditions and the flexibility of decision-making through the dynamic weight mechanism and feature fusion algorithm, providing a solid technical support for the intelligent management of cleaning and spraying data.

[0049] Embodiment 4: This embodiment involves the in-depth implementation of the dynamic weight calculation unit and the execution instruction generation module. The dynamic weight calculation unit introduces an adaptive adjustment mechanism when calculating the first dynamic decision factor to enhance the weight difference of high-entropy features. Specifically, when the hash entropy value of the optimized verification feature vector is greater than the entropy value of the initial decision benchmark vector, the system automatically increases the first decision coefficient from the initial value (such as 0.6) to 0.8, and vice versa to 0.4. For example, assume that the device operation pressure data corresponding to a certain feature vector fluctuates greatly in the time series, and its hash entropy value is significantly higher than the entropy value of the benchmark vector. At this time, the first decision coefficient increases, making the first dynamic decision factor of this feature vector increase, so that it occupies a higher proportion in the final weight, highlighting its influence on decision-making. The calculation of the second dynamic decision factor uses the standardized Euclidean distance. First, the feature vector is subjected to Z-score standardization processing to eliminate the influence of different feature dimensions (such as Pa for pressure and m³ / h for flow rate). Taking the device operation temperature (unit: °C) and spraying flow rate (unit: L / min) as an example, after standardization, both are converted into dimensionless values, and then the absolute value of the distance from the benchmark vector is calculated, multiplied by the second decision coefficient with dynamic adjustment (initial value 0.4, which can float within ±20% according to the overall fluctuation range of the data). For example, if the fluctuation range of the temperature feature suddenly increases during a certain period, the system automatically increases the second decision coefficient to 0.48 to enhance the role of the distance factor in weight calculation, ensuring that the weight allocation can adapt to the change of data distribution and improving the robustness of the calculation.

[0050] The execution instruction generation module inputs the data chain decision feature map into the instruction generator based on the smart contract, and the smart contract predefines the equipment control logic. Taking the common cleaning and spraying equipment combination as an example, when the spraying pressure characteristic value in the data chain decision feature map exceeds the preset threshold (such as 8MPa), the smart contract triggers the pressure adjustment instruction. The instruction generator calculates the specific parameters through the characteristic value and the preset control function (such as the linear mapping function). Assuming that the current characteristic value is 9MPa and the preset function is "instruction parameter = characteristic value - threshold", the generated pressure adjustment instruction parameter is 1MPa, that is, the control device reduces the pressure by 1MPa. The instruction generator adopts a modular design and supports control instruction templates of different types of equipment, which are automatically matched through the equipment type identification field in the feature map. For example, when the feature map carries the identification of "equipment type = high pressure cleaning machine", the instruction generator automatically calls the control template of the high pressure cleaning machine and generates instructions including pressure valve opening adjustment, cleaning liquid pump start and stop, etc.; if the identification is "equipment type = electrostatic spray gun", the control template of the spray gun is called to generate instructions such as spray flow adjustment and electrostatic voltage adjustment.

[0051] The generated device control instruction set contains instruction code, parameter value, execution timestamp and check code. Taking the pressure adjustment instruction of the high-pressure cleaner as an example, the instruction code "CMD_001" represents pressure adjustment, the parameter value "-1MPa" means reducing the pressure by 1MPa, and the execution timestamp "2025-05-30T14:30:00Z" specifies the instruction execution time. The check code is generated by hashing the instruction code, parameter value and timestamp through the SHA-256 algorithm to ensure the integrity of the instruction during transmission. The instruction is broadcast to the device control node through the blockchain. After the node receives the instruction, it first uses the check code to verify whether the instruction has been tampered with. If the verification passes, it parses the instruction parameters and executes the corresponding operation. For example, after the control node of the high-pressure cleaner parses the pressure adjustment instruction, it sends an electrical signal to the pressure valve actuator to adjust the valve opening to reduce the pressure, and monitors the pressure change in real time through the sensor, and feeds back the execution result to the cleaning parameter acquisition module to form a closed-loop control.

[0052] In the dynamic weight calculation process, the combination of adaptive adjustment mechanism and standardized processing can effectively cope with the characteristic changes under different working conditions. For example, when the equipment switches from conventional cleaning mode to fine spray mode, the volatility of the flow characteristics may increase significantly, and its hash entropy value will increase accordingly and exceed the reference vector. At this time, the first decision coefficient is automatically increased, which increases the weight of the flow characteristics. The system pays more attention to the changes in flow data when making decisions, thereby generating control instructions that better meet the needs of fine spraying. The application of standardized Euclidean distance avoids weight deviations caused by dimensional differences, ensuring that different characteristics such as temperature, pressure, and flow have a fair competitive basis in weight calculation.

[0053] The smart contract mechanism of the instruction generation module ensures the transparency and immutability of the control logic. For example, the logic of "automatically start the cooling system when the temperature exceeds 60°C" is preset in the smart contract. This logic is stored in the form of code on each node of the blockchain, and no single node can modify it without authorization. When the temperature feature value in the feature map reaches 65°C, the instruction generator automatically triggers the start instruction of the cooling system, and the instruction is synchronized to all nodes through the blockchain network to ensure the consistency and reliability of device control. In addition, the modular instruction template design improves the compatibility of the system and can quickly adapt to new types of cleaning and spraying equipment. Only by adding the corresponding control template to the system can the integrated management of new equipment be realized.

[0054] The dynamic weight calculation unit realizes the flexible adjustment of feature weights through the adaptive mechanism and the normalization method. The instruction generation module ensures the accuracy and compatibility of instructions with the help of smart contracts and modular design. This design not only improves the system's response ability to complex working conditions but also ensures the credibility and traceability of control instructions through the distributed characteristics of the blockchain, providing a complete technical solution for the intelligent and automated management of cleaning and spraying equipment. From the weight calculation of the feature vector to the generation and execution of instructions, each link is based on the actual data characteristics and device control requirements, forming a complete closed-loop from data collection, processing to decision execution, ensuring that the system can operate efficiently and reliably.

[0055] Embodiment 5: This embodiment involves the underlying implementation of spatio-temporal dimension division and feature sharding encryption. The spatio-temporal dimension division is realized through the dual mechanisms of timestamp slicing and spatial grid coding. Taking a certain spraying production line as an example, the equipment is equipped with a GPS positioning module and a high-precision clock to collect the real-time operation status data stream (such as pressure, flow rate, temperature, operation coordinates, etc.). The timestamp slicing can be set at an interval of 100 ms (adjustable according to the data update frequency), dividing the continuous data stream into multiple time windows, and each window corresponds to a data segment. For example, at the moment of 10:00:00 on May 30, 2025, the system generates a data segment with a timestamp of "20250530100000", containing the equipment operation data within 100 ms before and after this moment.

[0056] Spatial grid coding uses the grid division algorithm of the Geographic Information System (GIS) to divide the equipment operation area into grid cells of 50m×50m (the resolution can be adjusted according to the operation range). Suppose the current GPS coordinates of a certain equipment are (longitude 118.78°, latitude 31.95°), which are mapped to the grid cell numbered "A05" through the grid division algorithm. Each data segment needs to be appended with the corresponding grid code to form a spatio-temporal associated shard identifier, such as "20250530100000-001_A05", where "20250530100000" is the timestamp, "001" is the shard sequence number (indicating the first data shard within this time window), and "A05" is the spatial grid code. This identification method enables each data shard to have clear spatio-temporal coordinates, facilitating subsequent retrieval and verification of data by region or time range.

[0057] Feature shard encryption uses the asymmetric key pairs of distributed nodes to perform cross-signature verification on the shard set. Suppose there are three distributed nodes, node A, node B, and node C, in the blockchain network, and each node has a unique public-private key pair (public key Puk, private key Prk). When the equipment collects the data shard "20250530100000-001_A05", first the source node (such as node A) uses its own private key Prk_A to digitally sign the shard, generating the signature S1, and then transmits the shard and the signature S1 to the adjacent node B. After receiving the data shard, node B uses the public key Puk_A of node A to verify the validity of the signature S1: generating a digest of the shard content through the hash algorithm, and then decrypting the signature S1 with Puk_A to compare whether the two are consistent. If the verification passes, node B uses its own private key Prk_B to perform a secondary signature on the shard, generating the signature S2, and then transmits the shard, S1, and S2 to node C. Node C repeats the above verification process, verifying the signatures of node A and node B in turn. After the verification passes, it adds its own signature S3, finally forming a shard set containing multi-node cross-signatures.

[0058] The shard set after cross-signature is organized through the Merkle tree structure. The hash value of each shard contains the original data and all signature information. For example, if the shard content is D and the signatures are S1, S2, and S3, then the shard hash value H(D, S1, S2, S3) is generated through the SHA-256 algorithm. The bottom-level leaf nodes are the hash values of each shard, and adjacent leaf nodes are combined in pairs to generate the hash value of the upper-level parent node until the root hash value is generated. The root hash value is stored synchronously with the shard set in the blocks of each node in the blockchain. When it is necessary to verify the data integrity, only need to compare whether the root hash values stored in each node are the same; if it is necessary to trace the data source, the signature chain of each shard (S1→S2→S3) can be used to determine the sequence of nodes participating in the verification, ensuring that the data cannot be tampered with and the source can be traced.

[0059] Taking the cross-regional operation scenario of cleaning and spraying equipment as an example, assume that the equipment moves from grid A05 to grid B07. The corresponding data stream is sliced into multiple shards in chronological order, such as "20250530100000-001_A05", "20250530100000-002_B07", etc. Each shard is cross-signed and verified by adjacent nodes during the transmission process to ensure that the data generated when the equipment operates in different regions is reliably recorded. For example, when the equipment enters grid B07, the newly generated shard is cross-signed by nodes B, C, and D to form a new signature chain, reflecting the verification coverage of distributed nodes in the spatial dimension.

[0060] In the data retrieval scenario, if it is necessary to query the equipment operation data in grid A05 during the period from 10:00 to 10:01 on May 30, 2025, the corresponding shard set can be quickly located through the spatio-temporal identifier. The system filters out all shards that meet the conditions according to the timestamps from "20250530100000" to "20250530100100" and the grid code "A05", and then verifies the integrity of each shard through the hash path of the Merkle tree to ensure the accuracy of the retrieved data.

[0061] The combination of the cross-signature mechanism and the Merkle tree structure not only realizes the encrypted storage of data shards, but also enhances the security of the system through the consensus verification of distributed nodes. For example, if a malicious node attempts to tamper with the shard content, it must simultaneously tamper with the signature information of all participating signature nodes, which is almost impossible to achieve under the blockchain consensus mechanism. In addition, the spatio-temporal dimension division makes the data have a clear spatio-temporal index, facilitating data aggregation and analysis according to business requirements (such as regional operation statistics, time period performance analysis), and improving the data management efficiency of the system.

[0062] The processing flow of the entire embodiment closely revolves around the spatio-temporal attributes of data and the requirements of distributed verification. By slicing timestamps and spatial grid codes, clear spatio-temporal coordinates are assigned to the data, and asymmetric key cross-signature and Merkle tree structure are used to achieve data encryption and integrity verification. This design not only meets the data chunking requirements of cleaning and spraying equipment in a dynamic operation environment, but also guarantees the credibility and traceability of data through the distributed characteristics of the blockchain, providing a reliable technical foundation for cross-regional and cross-time data management and application. From the generation of data shards, signature verification to Merkle tree construction, each link is closely connected, forming a complete spatio-temporal data encryption and verification system to ensure that the system can efficiently manage cleaning and spraying data in a complex physical space and time series.

[0063] It should be noted that, in this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, such that a process, method, article or apparatus comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or apparatus.

[0064] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A blockchain-based cleaning and spraying data management system, characterized in that Including: A cleaning parameter acquisition module, configured to obtain the operation status data stream of the spraying device through distributed nodes; A data sharding and encryption module, configured to perform spatio-temporal dimension division and feature sharding encryption on the operation status data stream to generate a set of data verification feature matrices; A cross-chain feature verification module, configured to perform consensus mechanism screening and feature correlation verification on the set of data verification feature matrices to obtain a set of optimized verification feature matrices; A dynamic decision reconstruction module, configured to perform feature fusion based on a hash time lock on the set of optimized verification feature matrices to generate a data chain decision feature map; An execution instruction generation module, configured to generate a device control instruction set according to the data chain decision feature map.

2. The cleaning and spraying data management system based on blockchain according to claim 1, wherein The data sharding and encryption module includes: A data stream splitting unit, configured to dynamically split the operation status data stream according to a preset block size to form a set of data shards; A feature encryption unit, configured to input each shard in the set of data shards into a data shard encryptor based on a Merkle tree to generate the set of data verification feature matrices.

3. The blockchain-based cleaning and spraying data management system according to claim 2, wherein, The cross-chain feature verification module includes: A feature matrix deconstruction unit, configured to perform chain parsing on the data verification feature matrix along the time stamp dimension to obtain a set of feature verification vectors; A consensus association calculation unit, configured to calculate the consensus verification value between any two feature vectors in the set of feature verification vectors to generate a feature association consensus matrix; A dynamic screening unit, configured to adjust the weight of the feature association consensus matrix according to the hash difference between adjacent blocks in the set of feature verification vectors to generate a consensus constraint association matrix; A feature optimization unit, configured to perform cross-chain convolution aggregation on the consensus constraint association matrix and the set of feature verification vectors to obtain the optimized verification feature matrix.

4. The cleaning and spraying data management system based on blockchain according to claim 3, wherein The consensus association calculation unit includes: A feature mapping sub-unit, configured to map each feature vector in the set of feature verification vectors to an orthogonal hash space to obtain a set of mapped feature vectors; A consensus correlation analysis sub-unit, configured to calculate the hash similarity between any two mapped feature vectors in the set of mapped feature vectors to generate the feature association consensus matrix composed of multiple consensus association values.

5. The blockchain-based cleaning and spraying data management system according to claim 4, wherein The feature optimization unit is specifically implemented as: Performing chain convolution processing on the consensus constraint association matrix to generate a consensus constraint feature tensor; Inputting the set of feature verification vectors and the consensus constraint feature tensor into a bidirectional time stamp encoding network to generate a set of chain context feature vectors; Performing matrix reorganization on the set of chain context feature vectors to generate the optimized verification feature matrix.

6. The blockchain-based cleaning and spraying data management system according to claim 5, wherein, The dynamic decision reconstruction module includes: A feature matrix compression unit, configured to perform maximum pooling processing on each matrix in the set of optimized verification feature matrices along the time stamp dimension to generate a set of optimized verification feature vectors; A hash entropy value calculation unit, configured to calculate the hash entropy value of each feature vector in the set of optimized verification feature vectors to generate a dynamic hash entropy set; A reference feature determination unit, configured to select the optimized verification feature vector corresponding to the maximum hash entropy value in the dynamic hash entropy set as the initial decision reference vector; A dynamic weight calculation unit, configured to calculate the dynamic decision weights of each feature vector according to the hash distance between each feature vector in the set of optimized verification feature vectors and the initial decision reference vector and the hash entropy value of each feature vector, so as to generate a dynamic weight set; A chained reconstruction unit, configured to perform weighted aggregation on the set of optimized verification feature vectors by using the dynamic weight set to generate the data chained decision feature map.

7. The blockchain-based cleaning and spraying data management system according to claim 6, wherein The hash entropy value calculation unit is specifically implemented as: Calculating the mean vector and variance vector of the optimized verification feature vector; Performing element-by-element difference calculation on the optimized verification feature vector and the mean vector, and performing a square operation on the difference result to generate a feature hash difference vector; Calculating the overall median of the feature hash difference vector; performing a ratio operation on the median and the square value of the variance vector, and inputting the result into a normalization function to obtain the hash entropy value.

8. The blockchain-based cleaning and spraying data management system according to claim 7, wherein The dynamic weight calculation unit is specifically implemented as: Multiplying the hash entropy value of the optimized verification feature vector and the hash entropy value of the initial decision reference vector by a first decision coefficient to obtain a first dynamic decision factor; Multiplying the absolute value of the Euclidean distance between the optimized verification feature vector and the initial decision reference vector by a second decision coefficient to obtain a second dynamic decision factor; Performing weighted summation on the first dynamic decision factor and the second dynamic decision factor to obtain the dynamic decision weight.

9. The blockchain-based cleaning and spraying data management system according to claim 8, wherein The execution instruction generation module is specifically implemented as: Inputting the data chained decision feature map into an instruction generator based on a smart contract to generate the device control instruction set, and the instruction set is used to control the operating parameters of the cleaning and spraying device.

10. The blockchain-based cleaning and spraying data management system according to claim 1, characterized in that, The spatio-temporal dimension division includes: Segmenting the operation state data stream into continuous data segments according to time stamps, and attaching spatial grid encoding to each segment to form a spatio-temporal associated shard set; Performing cross-signature verification on the shard set by using the asymmetric key of the distributed node for the feature shard encryption.

Images