A blockchain-based environmental monitoring data archiving method

Abstract

The application provides an environment monitoring data storage method based on a block chain, and relates to the technical field of data processing.In the application, firstly, target environment monitoring data is subjected to matrix processing to form an environment data matrix sequence; secondly, the environment data matrix sequence is subjected to multi-angle semantic mining to form a plurality of semantic mining vectors, and based on the plurality of semantic mining vectors, an environment data global vector is formed; then, based on the environment data global vector, abnormality identification is performed to form an environment abnormality analysis result; on one hand, when the environment abnormality analysis result reflects that there is abnormality, the encrypted environment monitoring data and the environment abnormality analysis result are stored into a target block chain; on the other hand, when the environment abnormality analysis result reflects that there is no abnormality, the target environment monitoring data and the environment abnormality analysis result are stored into the target block chain.Based on the above method, the problem that data privacy and accessibility are difficult to effectively balance in the prior art can be improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically, to a method for storing environmental monitoring data based on blockchain. Background Technology

[0002] Blockchain is a novel application technology of computer technology, encompassing distributed data storage, peer-to-peer transmission, consensus protocols, and encryption algorithms. In a blockchain system, data blocks are sequentially linked to form a chain-like data structure, and a distributed ledger is cryptographically guaranteed to be immutable and unforgeable. Furthermore, due to its decentralized, immutable, and autonomous characteristics, blockchain applications are becoming increasingly widespread. For example, storing environmental monitoring data via blockchain can effectively ensure data reliability and facilitate reliable traceability. However, current technologies typically encrypt environmental monitoring data before storage, leading to low accessibility; conversely, leaving the data unencrypted compromises data privacy. Therefore, existing technologies face the challenge of effectively balancing data privacy and accessibility. Summary of the Invention

[0003] In view of this, the purpose of this invention is to provide a blockchain-based method for storing environmental monitoring data, so as to improve the problem of difficulty in effectively balancing data privacy and accessibility in the prior art.

[0004] To achieve the above objectives, the embodiments of the present invention adopt the following technical solutions:

[0005] A blockchain-based method for storing environmental monitoring data includes:

[0006] The target environmental monitoring data is matrixed to form an environmental data matrix sequence, wherein each environmental data matrix in the environmental data matrix sequence includes environmental data obtained by multiple environmental monitoring devices within the same time segment;

[0007] Multi-angle semantic mining is performed on the environmental data matrix sequence to form multiple semantic mining vectors, and the multiple semantic mining vectors are fused to form a global environmental data vector, wherein the global environmental data vector is used to characterize the global semantic information of the target environmental monitoring data;

[0008] Anomaly identification is performed based on the global vector of the environmental data to form an environmental anomaly analysis result of the target environmental monitoring data.

[0009] When the environmental anomaly analysis results indicate that the target environmental monitoring data is abnormal, the target environmental monitoring data is encrypted to form encrypted environmental monitoring data, and the encrypted environmental monitoring data and the environmental anomaly analysis results are stored in the target blockchain.

[0010] When the environmental anomaly analysis results indicate that the target environmental monitoring data does not exhibit any anomalies, the target environmental monitoring data and the environmental anomaly analysis results are stored in the target blockchain.

[0011] In some preferred embodiments, in the above-described blockchain-based environmental monitoring data storage method, the step of performing multi-angle semantic mining on the environmental data matrix sequence to form multiple semantic mining vectors, and fusing the multiple semantic mining vectors to form a global environmental data vector, includes:

[0012] In the environmental data matrix sequence, A environmental data matrices are determined;

[0013] The A environmental data matrices are subjected to first semantic mining to form a first environmental vector combination;

[0014] The A environmental data matrices are subjected to second semantic mining to form a second environmental vector combination;

[0015] Candidate screening indicators are determined, and A screening parameters with a sequential relationship among the candidate screening indicators are assigned values ​​to form a first screening parameter distribution. Based on the first screening parameter distribution, a second screening parameter distribution is generated. The assignment results include a first value and a second value. The first value is used to reflect the corresponding vector being selected, and the second value is used to reflect the corresponding vector being discarded. The assignment results of the screening parameters at corresponding coordinates in the first screening parameter distribution and the second screening parameter distribution are different.

[0016] The first environmental vector combination is filtered based on the first filtering parameter distribution to form a first filtering vector combination, and the second environmental vector combination is filtered based on the second filtering parameter distribution to form a second filtering vector combination, and a target filtering vector combination is formed based on the first filtering vector combination and the second filtering vector combination.

[0017] By combining the target filtering vector combination, the first environment vector combination, and the second environment vector combination, a global environment data vector is determined.

[0018] In some preferred embodiments, in the above-described blockchain-based environmental monitoring data storage method, the step of performing first semantic mining on the A environmental data matrices to form a first environmental vector combination includes:

[0019] Focus mining is performed on any one of the A environmental data matrices to form a first environmental focus vector corresponding to the environmental data matrix. The first environmental focus vector focuses on reflecting the environmental semantic information of the first focused part in the environmental data matrix.

[0020] Based on the first environmental focus vector, the environmental data matrix is ​​focused and transferred to form a transferred environmental data matrix corresponding to the environmental data matrix;

[0021] The transfer environment data matrix is ​​subjected to focused mining to form a second environment focused vector corresponding to the transfer environment data matrix. The second environment focused vector focuses on reflecting the environmental semantic information of the second focused part in the transfer environment data matrix. The second focused part is different from the first focused part.

[0022] The first environment focus vector is deeply mined to form a first environment depth vector, and the second environment focus vector is deeply mined to form a second environment depth vector;

[0023] The first environment depth vector and the second environment depth vector are merged to form a first environment vector, and the first environment vectors corresponding to the A environment data matrices are combined to form a first environment vector combination.

[0024] In some preferred embodiments, in the above-described blockchain-based environmental monitoring data storage method, the step of performing focused mining on any one of the A environmental data matrices to form a first environmental focus vector corresponding to the environmental data matrix includes:

[0025] Perform a first convolution operation and a second convolution operation on any one of the A environmental data matrices to form a first environmental convolution vector and a second environmental convolution vector, wherein the first environmental convolution vector and the second environmental convolution vector have the same size.

[0026] Based on one of the first environment convolution vector and the second environment convolution vector, cross-attention processing or gating adjustment is performed on the other convolution vector to form the first environment focus vector corresponding to the environment data matrix.

[0027] In some preferred embodiments, in the above-described blockchain-based environmental monitoring data storage method, the step of focusing and transferring the environmental data matrix according to the first environmental focus vector to form a transferred environmental data matrix corresponding to the environmental data matrix includes:

[0028] Based on the first environmental focus vector, a first focus parameter distribution corresponding to the environmental data matrix is ​​determined, and based on the first focus parameter distribution, a second focus parameter distribution is determined, wherein the first focus parameter distribution is used to reflect the distribution position of the first focus portion, and the second focus parameter distribution is used to reflect the distribution position of the second focus portion.

[0029] Based on the second focusing parameter distribution and the environmental data matrix, the transfer environmental data matrix is ​​determined, wherein the first value in the second focusing parameter distribution is used to reflect that the environmental data at the corresponding distribution location belongs to the second focusing part, and the second value in the second focusing parameter distribution is used to reflect that the environmental data at the corresponding distribution location does not belong to the second focusing part.

[0030] In some preferred embodiments, in the above-described blockchain-based environmental monitoring data storage method, the steps of determining the first focusing parameter distribution corresponding to the environmental data matrix based on the first environmental focusing vector, and determining the second focusing parameter distribution based on the first focusing parameter distribution, include:

[0031] Without changing the vector size, the first environment focusing vector is linearly mapped to form a first environment mapping vector, and the first environment mapping vector is nonlinearly activated to form a first environment activation vector.

[0032] The mean value of each vector parameter in the first environment activation vector is calculated to obtain the mean value of the vector parameters. Based on the mean value of the vector parameters, the first environment activation vector is binarized and mapped to form the first focusing parameter distribution corresponding to the environment data matrix.

[0033] The difference between the first value and each parameter in the first focusing parameter distribution is calculated to form the second focusing parameter distribution.

[0034] In some preferred embodiments, in the above-described blockchain-based environmental monitoring data storage method, the first environmental vector combination includes the first environmental vector corresponding to each of the A environmental data matrices, and the second environmental vector combination includes the second environmental vector corresponding to each of the A environmental data matrices.

[0035] The step of determining the global vector of environmental data by fusing the target filtering vector combination, the first environmental vector combination, and the second environmental vector combination includes:

[0036] In the first environment vector combination and the second environment vector combination, the first environment vector and the second environment vector corresponding to the same environment data matrix are respectively subjected to vector parameter mixing processing to form the mixed environment vectors corresponding to the A environment data matrices.

[0037] Based on the hybrid environment vectors corresponding to each of the A environmental data matrices, a hybrid environment vector combination is determined;

[0038] Based on the target filtering vector combination and the mixed environment vector combination, a global vector of environmental data is determined.

[0039] In some preferred embodiments, in the above-described blockchain-based environmental monitoring data storage method, the A environmental data matrices include a first environmental data matrix;

[0040] The step of mixing the vector parameters of the first environment vector and the second environment vector corresponding to the same environment data matrix in the first environment vector combination and the second environment vector combination to form the mixed environment vector corresponding to each of the A environment data matrices includes:

[0041] The first environment vector corresponding to the first environment data matrix is ​​determined from the first combination of environment vectors, and the second environment vector corresponding to the first environment data matrix is ​​determined from the second combination of environment vectors.

[0042] Determine the first weight parameter and the second weight parameter;

[0043] Based on the first weight parameter, the parameters of the first environment vector corresponding to the first environment data matrix are adjusted to form the first environment adjustment vector;

[0044] Based on the second weight parameter, the parameters of the second environment vector corresponding to the first environment data matrix are adjusted to form the second environment adjustment vector;

[0045] The first environment adjustment vector and the second environment adjustment vector are mixed by vector parameter processing to form a mixed environment vector corresponding to the first environment data matrix.

[0046] In some preferred embodiments, in the above-described blockchain-based environmental monitoring data storage method, the step of mixing the vector parameters of the first environmental adjustment vector and the second environmental adjustment vector to form a mixed environmental vector corresponding to the first environmental data matrix includes:

[0047] A first local vector is determined in the first environment adjustment vector, and a second local vector is determined in the second environment adjustment vector based on the distribution position of the first local vector in the first environment adjustment vector, wherein the distribution position of the second local vector in the second environment adjustment vector is completely different from the distribution position of the first local vector in the first environment adjustment vector.

[0048] The first local vector and the second local vector are concatenated to form the hybrid environment vector corresponding to the first environment data matrix.

[0049] In some preferred embodiments, in the above-described blockchain-based environmental monitoring data notarization method, the step of identifying anomalies based on the global vector of the environmental data to form an environmental anomaly analysis result of the target environmental monitoring data includes:

[0050] The global vector of environmental data is processed by full connection to form a fully connected vector of environmental data;

[0051] Based on the target classification function, the fully connected vector of the environmental data is activated to form a target probability distribution, wherein the target probability distribution includes the probability that the target environmental monitoring data does not have anomalies and the probability that it has anomalies;

[0052] Based on the target probability distribution, the environmental anomaly analysis results of the target environmental monitoring data are determined.

[0053] This invention provides a blockchain-based method for storing environmental monitoring data. First, the target environmental monitoring data is matrixed to form an environmental data matrix sequence. Second, multi-angle semantic mining is performed on the environmental data matrix sequence to form multiple semantic mining vectors. Based on these vectors, a global environmental data vector is formed. Then, anomaly identification is performed based on the global environmental data vector to generate an environmental anomaly analysis result. On one hand, when the environmental anomaly analysis result indicates the presence of anomalies, the encrypted environmental monitoring data and the environmental anomaly analysis result are stored in the target blockchain. On the other hand, when the environmental anomaly analysis result indicates the absence of anomalies, the target environmental monitoring data and the environmental anomaly analysis result are also stored in the target blockchain. Based on the above method, different storage schemes are used depending on whether anomalies are present, which allows for a balance between data privacy and accessibility. In addition, since the basis for anomaly identification (i.e., the global vector of environmental data) is formed by matrix processing and multi-angle semantic mining of the target environmental monitoring data, the semantic representation accuracy of the identification basis will be high (multi-angle semantic mining after matrix processing can facilitate the discovery of correlations between environmental data from different environmental monitoring devices). Thus, the reliability of environmental anomaly analysis results can be guaranteed, thereby making the corresponding storage more reliable and improving the problem of effectively balancing data privacy and accessibility in existing technologies.

[0054] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0055] Figure 1 This is a structural block diagram of an electronic device provided in an embodiment of the present invention.

[0056] Figure 2 The flowchart illustrates the steps of the blockchain-based environmental monitoring data storage method provided in this embodiment of the invention.

[0057] Figure 3 This is a schematic diagram of an environmental data matrix provided in an embodiment of the present invention.

[0058] Figure 4 This is a schematic diagram illustrating multi-angle semantic mining and fusion provided in an embodiment of the present invention.

[0059] Figure 5 This is a schematic diagram of the first semantic mining method provided in an embodiment of the present invention.

[0060] Figure 6 This is a schematic diagram illustrating the determination of the focusing parameter distribution provided in an embodiment of the present invention. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0062] like Figure 1 As shown, an embodiment of the present invention provides an electronic device. The electronic device may include a memory and a processor.

[0063] In detail, the memory and the processor are electrically connected directly or indirectly to enable data transmission or interaction. For example, they can be electrically connected via one or more communication buses or signal lines. The memory may store at least one software functional module (computer program) that exists in the form of software or firmware. The processor can be used to execute the executable computer program stored in the memory, thereby implementing the blockchain-based environmental monitoring data storage method provided in the embodiments of the present invention (as described below).

[0064] Optionally, the memory may be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc. The processor may be a general-purpose processor, including a Central Processing Unit (CPU), Network Processor (NP), System on Chip (SoC), etc.; it may also be a Digital Signal Processor (DSP), Application-Specific Integrated Circuit (ASIC), Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0065] and, Figure 1 The structure shown is for illustrative purposes only; the electronic device may also include components that are more advanced than those shown. Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown may include, for example, a communication unit for exchanging information with other devices. In an alternative example, the electronic device may be a server with data processing capabilities.

[0066] Combination Figure 2 This invention also provides a blockchain-based method for storing environmental monitoring data, which can be applied to the aforementioned electronic device. The method steps defined in the process of the blockchain-based environmental monitoring data storage method can be implemented by the electronic device.

[0067] The following will be about Figure 2 The specific process shown will be explained in detail.

[0068] Step S110: Perform matrix processing on the target environmental monitoring data to form an environmental data matrix sequence.

[0069] In this embodiment of the invention, the electronic device can perform matrix processing on the target environmental monitoring data to form an environmental data matrix sequence. Each environmental data matrix in the sequence includes environmental data monitored by multiple environmental monitoring devices within the same time segment. For example, for time segment 1, the environmental data from each monitoring device at each moment within time segment 1 can be matrixed to form environmental data matrix 1. Similarly, for time segment 2, the environmental data from each monitoring device at each moment within time segment 2 can be matrixed to form environmental data matrix 2. Figure 3 As shown, in any environmental data matrix, the data in the same column includes environmental data from various environmental monitoring devices at the same time, and the data in the same row includes environmental data from one environmental monitoring device at various times within a time segment. Based on this, in the subsequent semantic mining process, correlation mining can be performed on related semantic information in both the time dimension and the spatial dimension. Furthermore, it should be noted that the device types of the multiple environmental monitoring devices can be the same, partially the same, or completely different. For example, device types can include, but are not limited to, temperature sensors, humidity sensors, noise sensors, PM10 detectors, PM2.5 detectors, NO2 detectors, SO2 detectors, and O3 detectors. It should also be noted that the multiple environmental monitoring devices can be deployed in a single space (e.g., with a distribution distance less than a threshold, which can be set according to actual needs; the smaller the threshold, the stronger the spatial correlation), thus ensuring that the multiple environmental monitoring devices are spatially correlated.

[0070] Step S120: Perform multi-angle semantic mining on the environmental data matrix sequence to form multiple semantic mining vectors, and fuse the multiple semantic mining vectors to form a global environmental data vector.

[0071] In this embodiment of the invention, after forming the environmental data matrix sequence, the electronic device can perform multi-angle semantic mining on the environmental data matrix sequence to form multiple semantic mining vectors, and fuse these multiple semantic mining vectors to form a global environmental data vector. The global environmental data vector is used to represent the global semantic information of the target environmental monitoring data. That is, by performing different semantic mining operations, different semantic information can be extracted. Then, by fusing the semantic information, global semantic information is obtained, thus achieving a full representation of the semantic information in the environmental data matrix sequence.

[0072] Step S130: Based on the global vector of the environmental data, perform anomaly identification to form an environmental anomaly analysis result of the target environmental monitoring data.

[0073] In this embodiment of the invention, after forming the global vector of environmental data, the electronic device can perform anomaly identification based on the global vector of environmental data to form an environmental anomaly analysis result of the target environmental monitoring data. Exemplarily, steps S120 and S130 can be implemented using a neural network model. For example, the neural network model can include an encoding network and a decoding network. The encoding network can be used to execute step S120, that is, to encode the environmental data matrix sequence to achieve semantic information mining. The decoding network can be used to execute step S130, that is, to decode the global vector of environmental data to achieve anomaly identification. Based on this, environmental anomalies in the corresponding space can be monitored globally and holistically, avoiding the problem of unreliable results caused by data analysis and identification from individual environmental monitoring devices. For example, when environmental pollution exists, local data is difficult to accurately characterize environmental pollution, but by adopting this global approach, various local data complement each other, ensuring the reliability of the results.

[0074] Step S140: When the environmental anomaly analysis result reflects that the target environmental monitoring data is abnormal, the target environmental monitoring data is encrypted to form encrypted environmental monitoring data, and the encrypted environmental monitoring data and the environmental anomaly analysis result are stored in the target blockchain.

[0075] In this embodiment of the invention, after generating the environmental anomaly analysis result, the electronic device can, when the environmental anomaly analysis result reflects an anomaly in the target environmental monitoring data, encrypt the target environmental monitoring data (any existing encryption technology can be used, without specific limitations), generating encrypted environmental monitoring data, and then store the encrypted environmental monitoring data and the environmental anomaly analysis result in the target blockchain. That is, when storing anomalies, the target environmental monitoring data can be encrypted to ensure data privacy; however, the environmental anomaly analysis result is not encrypted to facilitate access and viewing.

[0076] Step S150: When the environmental anomaly analysis result reflects that the target environmental monitoring data does not have any anomalies, the target environmental monitoring data and the environmental anomaly analysis result are stored in the target blockchain.

[0077] In this embodiment of the invention, after generating the environmental anomaly analysis result, the electronic device can store the target environmental monitoring data and the environmental anomaly analysis result in the target blockchain when the environmental anomaly analysis result reflects that there is no anomaly in the target environmental monitoring data. That is, when there is no anomaly, considering the relatively low value of the data, it can be directly stored for easy access and viewing. Furthermore, it should be noted that the specific process of data storage via blockchain can refer to relevant existing technologies and is not specifically limited here. Also, the target environmental monitoring data and the environmental anomaly analysis result can be stored on the same blockchain or on different blockchains.

[0078] Based on the above method, different storage schemes are used depending on whether anomalies are present, which allows for a balance between data privacy and accessibility. In addition, since the basis for anomaly identification (i.e., the global vector of environmental data) is formed by matrix processing and multi-angle semantic mining of the target environmental monitoring data, the semantic representation accuracy of the identification basis will be high (multi-angle semantic mining after matrix processing can facilitate the discovery of correlations between environmental data from different environmental monitoring devices). Thus, the reliability of environmental anomaly analysis results can be guaranteed, thereby making the corresponding storage more reliable and improving the problem of effectively balancing data privacy and accessibility in existing technologies.

[0079] In the first part, regarding step S120, it should be noted that the specific method for performing multi-angle semantic mining and fusion on the environmental data matrix sequence is not limited and can be configured according to the relevant needs in actual applications.

[0080] For example, in one feasible implementation, considering that there may be multiple types of data in the environmental data matrix sequence, each environmental data matrix can be preprocessed (such as normalization and standardization), and then different convolution processes can be performed to obtain different convolution vectors, that is, multiple semantic mining vectors. Then, the multiple semantic mining vectors can be concatenated or added to achieve fusion, thereby obtaining a global environmental data vector. Alternatively, the fusion result can be subjected to self-attention processing to obtain a global environmental data vector.

[0081] For example, in another possible implementation, in order to ensure that the formed global vector of environmental data has a high semantic representation capability, that is, to be able to fully represent the semantic information in the sequence of environmental data matrix, the above step S120 may further include steps S121, S122, S123, S124, S125 and S126, the specific contents of which are as follows.

[0082] Step S121: In the environmental data matrix sequence, determine A environmental data matrices.

[0083] In this embodiment of the invention, A environmental data matrices can be determined from the environmental data matrix sequence. It should be noted that the A environmental data matrices can be multiple environmental data matrices. Furthermore, the A environmental data matrices can be either a portion of the environmental data matrices in the environmental data matrix sequence (e.g., obtained by sampling at a certain sampling rate), or all the environmental data matrices in the environmental data matrix sequence, combined with… Figure 4 As shown.

[0084] Step S122: Perform first semantic mining on the A environmental data matrices to form a first environmental vector combination.

[0085] In this embodiment of the invention, after determining the A environmental data matrices, the A environmental data matrices can be subjected to first semantic mining to form a first environmental vector combination. For example, each environmental data matrix can be subjected to first semantic mining to obtain each first environmental vector, and then they can be combined together, such as arranged in order, to form a first environmental vector combination.

[0086] Step S123: Perform second semantic mining on the A environmental data matrices to form a second environmental vector combination.

[0087] In this embodiment of the invention, after determining the A environmental data matrices, the A environmental data matrices can be subjected to second semantic mining to form a second environmental vector combination. For example, the second semantic mining can be performed on each environmental data matrix separately to obtain each second environmental vector, which can then be combined together, such as arranged in order, to form a second environmental vector combination. It should be noted that the mining process for the second semantic mining and the first semantic mining can be the same or different. In this embodiment of the invention, the mining process can be the same, but the parameters of the corresponding neural network model can be different (specifically formed during the training of the neural network model). For example, the above-mentioned encoding network can include a first semantic mining unit and a second semantic mining unit. The first mining unit is used to execute step S122 to realize the first semantic mining; the second mining unit is used to execute step S123 to realize the second semantic mining. The first semantic mining unit and the second semantic mining unit can have the same network architecture (the network parameters can be learned and trained based on the corresponding sample environment monitoring data and labels representing whether anomalies exist; the training process can refer to relevant existing technologies, and is not specifically limited here). This ensures a high degree of fit between the first environment vector combination and the second environment vector combination, facilitating fusion in subsequent steps. Based on this, the specific implementation of the second semantic mining can be found in the explanation of the first semantic mining described later.

[0088] Step S124: Determine the candidate screening indicators, and assign values ​​to A screening parameters that have a sequential relationship among the candidate screening indicators to form a first screening parameter distribution, and generate a second screening parameter distribution based on the first screening parameter distribution.

[0089] In this embodiment of the invention, candidate screening indicators can be determined, and A screening parameters (corresponding to A environmental data matrices) with a sequential relationship among the candidate screening indicators can be assigned values ​​to form a first screening parameter distribution. Based on the first screening parameter distribution, a second screening parameter distribution is generated. The assignment results include a first value and a second value. The first value reflects the selection of the corresponding vector, and the second value reflects the rejection of the corresponding vector. The assignment results of the screening parameters at corresponding coordinates in the first and second screening parameter distributions are different. For example, the first value can be equal to 1, and the second value can be equal to 0. Thus, when the screening parameter at the first coordinate in the first screening parameter distribution is equal to 1, the screening parameter at the second coordinate in the second screening parameter distribution is equal to 0; and when the screening parameter at the second coordinate in the first screening parameter distribution is equal to 0, the screening parameter at the second coordinate in the second screening parameter distribution is equal to 1.

[0090] Step S125: Filter the first environment vector combination based on the first filtering parameter distribution to form a first filtering vector combination; filter the second environment vector combination based on the second filtering parameter distribution to form a second filtering vector combination; and form a target filtering vector combination based on the first filtering vector combination and the second filtering vector combination.

[0091] In this embodiment of the invention, after forming the first screening parameter distribution and the second screening parameter distribution, the first environmental vector combination can be filtered based on the first screening parameter distribution to form a first screening vector combination, and the second environmental vector combination can be filtered based on the second screening parameter distribution to form a second screening vector combination. Furthermore, a target screening vector combination is formed based on the first screening vector combination and the second screening vector combination. For example, when merging the first screening vector combination and the second screening vector combination, they can be mixed according to their corresponding order (in other embodiments, the second screening vector combination can be arranged after the first screening vector combination). For instance, the first screening vector combination and the second screening vector combination can be arranged sequentially according to the position of the environmental data matrix corresponding to each first screening vector in the first screening vector combination within the environmental data matrix sequence, and the position of the environmental data matrix corresponding to each second screening vector in the second screening vector combination within the environmental data matrix sequence, to form the target screening vector combination.

[0092] Step S126: Combine the target filtering vector combination, the first environment vector combination, and the second environment vector combination to determine the global environment data vector.

[0093] In this embodiment of the invention, after forming the target filtering vector combination, the target filtering vector combination, the first environment vector combination, and the second environment vector combination can be fused to determine the global environment data vector. It should be noted that the target filtering vector combination, the first environment vector combination, and the second environment vector combination all include A vectors.

[0094] Optionally, in step S122 above, the specific method of performing the first semantic mining on the A environmental data matrices is not limited. For example, in one feasible implementation, in order to enrich the semantic information expression dimension of the first environmental vector in the first environmental vector combination, so as to give full attention to different semantic information in the environmental data matrix and thus improve the accuracy of semantic representation, step S122 above may further include steps S122a, S122b, S122c, S122d, and S122e, the specific contents of which are as follows (in conjunction with...). Figure 5 (As shown).

[0095] Step S122a: Focus mining is performed on any one of the A environmental data matrices to form the first environmental focus vector corresponding to the environmental data matrix.

[0096] In this embodiment of the invention, any one of the A environmental data matrices (each environmental data matrix is ​​processed in the same way) can be subjected to focused mining to form a first environmental focus vector corresponding to the environmental data matrix. The first environmental focus vector emphasizes the environmental semantic information of the first focused portion of the environmental data matrix. In other words, focused mining allows for a focus on the environmental semantic information of the first focused portion of the environmental data matrix, thus achieving the capture of important semantic information.

[0097] Step S122b: Based on the first environmental focus vector, the environmental data matrix is ​​focused and transferred to form a transferred environmental data matrix corresponding to the environmental data matrix.

[0098] In this embodiment of the invention, after forming the first environment focus vector, the environment data matrix can be focused and transferred based on the first environment focus vector to form a transferred environment data matrix corresponding to the environment data matrix. That is, since the first environment focus vector focuses on the semantic information of the first focused portion, although the semantic information of other parts of the environment data matrix is ​​not as important as the first focused portion, it still has a certain representational function and therefore needs to be captured and mined. Based on this, the first environment focus vector can be used to focus and transfer the environment data matrix to form a transferred environment data matrix, thereby facilitating the focused attention and mining of other parts.

[0099] Step S122c: Focus mining is performed on the transfer environment data matrix to form a second environment focus vector corresponding to the transfer environment data matrix.

[0100] In this embodiment of the invention, after forming the transfer environment data matrix, the transfer environment data matrix can be subjected to focused mining to form a second environment focused vector corresponding to the transfer environment data matrix. The second environment focused vector focuses on reflecting the environmental semantic information of a second focused portion in the transfer environment data matrix, which is different from the first focused portion. It should be noted that the method of focused mining of the transfer environment data matrix can be different from or the same as the method of focused mining of the environment data matrix. In this embodiment of the invention, they can be the same. This ensures the compatibility between the first and second environment focused vectors, facilitating fusion after subsequent deep mining to improve the semantic richness of the fused first environment vector.

[0101] Step S122d: Deeply mine the first environment focus vector to form a first environment depth vector, and deeply mine the second environment focus vector to form a second environment depth vector.

[0102] In this embodiment of the invention, after forming the first environment focus vector and the second environment focus vector, the first environment focus vector can be deeply mined to form a first environment depth vector, and the second environment focus vector can be deeply mined to form a second environment depth vector. It should be noted that deep mining can include convolution (e.g., 64 convolution kernels, 3*3 kernel size, stride 1), pooling (e.g., a 2×2 pooling window, stride 2, activation function ReLU (Rectified Linear Unit)), and fully connected processing (e.g., output dimension 512). In this way, high-level, abstract semantic features can be extracted from the first and second environment focus vectors, thereby further improving the semantic representation capability.

[0103] Step S122e: Merge the first environment depth vector and the second environment depth vector to form a first environment vector, and merge the first environment vectors corresponding to the A environment data matrices to form a first environment vector combination.

[0104] In this embodiment of the invention, after forming the first environment depth vector and the second environment depth vector, the first environment depth vector and the second environment depth vector can be fused to form a first environment vector. Furthermore, based on the first environment vectors corresponding to each of the A environment data matrices, they can be merged to form a first environment vector combination. The specific fusion methods include, but are not limited to, any one of the following: splicing, addition, gating adjustment, cross-attention, etc. Additionally, after forming the first environment vector corresponding to each environment data matrix, the first environment vectors can be merged (permuted and combined) based on the arrangement relationship between the environment data matrices, thereby forming a first environment vector combination including A first environment vectors.

[0105] Optionally, in step S122a above, the specific method of focusing and mining any one of the A environmental data matrices is not limited. For example, in one feasible implementation, in order to fully extract the important semantic information in the environmental data matrix through focused mining, step S122a above may further include the following:

[0106] The first step is to perform a first convolution operation and a second convolution operation on any one of the A environment data matrices (for example, it can be implemented through a first convolutional network layer and a second convolutional network layer respectively) to form a first environment convolution vector and a second environment convolution vector, wherein the first environment convolution vector and the second environment convolution vector have the same size.

[0107] The second step involves applying cross-attention processing or gating adjustment to the other convolutional vector based on one of the first and second environment convolutional vectors to form the first environment focus vector corresponding to the environment data matrix. For example, the first environment convolutional vector can be mapped to a query vector based on the query matrix in the attention network layer, and the second environment convolutional vector can be mapped to a key vector and a value vector based on the key and value matrices in the attention network layer. Then, the value vector can be weighted and summed based on the attention parameters between the query vector and the key vector to obtain the first environment focus vector. Alternatively, the first environment convolutional vector can be linearly mapped (without changing the vector size) and non-linearly activated (e.g., using functions like sigmoid). The result can then be multiplied bitwise with the second environment convolutional vector to obtain the first environment focus vector.

[0108] Optionally, in step S122b above, the specific method of focusing and transferring the environmental data matrix is ​​not limited. For example, in one feasible implementation, in order to achieve attention to different semantic information through focusing and transferring, step S122b above may further include steps b1 and b2, the specific contents of which are as follows.

[0109] Step b1: Based on the first environmental focusing vector, determine the first focusing parameter distribution corresponding to the environmental data matrix, and based on the first focusing parameter distribution, determine the second focusing parameter distribution.

[0110] In this embodiment of the invention, a first focusing parameter distribution corresponding to the environmental data matrix can be determined based on the first environmental focusing vector, and a second focusing parameter distribution can be determined based on the first focusing parameter distribution. The first focusing parameter distribution reflects the distribution position of the first focusing portion (for example, in the first environmental focusing vector, a larger parameter value indicates greater importance, i.e., belonging to the first focusing portion), and the second focusing parameter distribution reflects the distribution position of the second focusing portion (for example, since the second focusing portion is different from the first focusing portion, a larger parameter in the first focusing parameter distribution corresponds to a smaller parameter in the second focusing parameter distribution).

[0111] Step b2: Determine the transfer environment data matrix based on the second focusing parameter distribution and the environment data matrix.

[0112] In this embodiment of the invention, after determining the second focusing parameter distribution, the transferred environmental data matrix can be determined based on the second focusing parameter distribution and the environmental data matrix. The first value in the second focusing parameter distribution reflects that the environmental data at the corresponding distribution location belongs to the second focusing portion, and the second value reflects that the environmental data at the corresponding distribution location does not belong to the second focusing portion. For example, when each parameter in the second focusing parameter distribution is equal to 1 or 0, and the second focusing parameter distribution and the environmental data matrix have the same size, the second focusing parameter distribution and the environmental data matrix can be multiplied bitwise to obtain the transferred environmental data matrix. This allows the environmental data corresponding to the semantic information of the first environmental focusing vector to be hidden, enabling subsequent focusing mining processes to focus on the semantic mining of other parts of the environmental data.

[0113] Optionally, in step b1 above, the specific method of determining the first and second focusing parameter distributions is not limited. For example, in one feasible implementation, in order to ensure that the determined first and second focusing parameter distributions can effectively characterize different parts of interest (i.e., effectively distinguish the first focusing part and the second focusing part), step b1 above may further include the following (in conjunction with...). Figure 6 (as shown)

[0114] The first step involves linearly mapping the first environment focus vector (e.g., y = Ax + b, where A and b are parameters of the linear mapping) without changing the vector size. It should be noted that maintaining the same vector size is based on the premise that the first environment focus vector and the environment data matrix have the same size. Therefore, if the first environment focus vector and the environment data matrix have different sizes, the size of the first environment focus vector can be adjusted using linear mapping. Specifically, this can be achieved by configuring the size of the weight matrix A so that the size of the resulting first environment mapping vector is equal to the size of the environment data matrix. This forms the first environment mapping vector, which is then non-linearly activated (e.g., using functions like sigmoid) to form the first environment activation vector.

[0115] The second step involves calculating the mean of each vector parameter in the first environment activation vector to obtain the mean value of the vector parameters. Based on the mean value of the vector parameters, the first environment activation vector is binarized to form a first focusing parameter distribution corresponding to the environment data matrix. (For example, for each vector parameter in the first environment activation vector, if the vector parameter is less than the mean value of the vector parameter, then the parameter at the corresponding coordinate in the first focusing parameter distribution is equal to 0.) Figure 6 The black areas in the diagram represent areas that are not under special attention, i.e., not part of the first focus area. If the vector parameter is greater than or equal to the mean of the vector parameter, then the parameter at the corresponding coordinate in the first focus parameter distribution is equal to 1. Figure 6 The white area in the image indicates areas that have received special attention, i.e., the first area of ​​focus.

[0116] The third step is to calculate the difference between the first value and each parameter in the first focusing parameter distribution to form the second focusing parameter distribution. The first value can be equal to 1. That is, the parameters in the first focusing parameter distribution that are equal to 1 can be updated to 0, and the parameters that are equal to 0 can be updated to 1, thereby filtering out the second focusing part that does not belong to the first focusing part.

[0117] Optionally, in step S126 above, the specific method of fusing the target filtering vector combination, the first environment vector combination, and the second environment vector combination is not limited. For example, in one feasible implementation, in order to achieve full fusion of different semantic information, step S126 above may further include steps S126a, S126b, and S126c (in other implementations, the vectors may be concatenated or added to obtain a global environment data vector), the specific content of which is as follows.

[0118] Step S126a: In the first environment vector combination and the second environment vector combination, the first environment vector and the second environment vector corresponding to the same environment data matrix are respectively subjected to vector parameter mixing processing to form the mixed environment vector corresponding to each of the A environment data matrices.

[0119] In this embodiment of the invention, in the first environment vector combination and the second environment vector combination, the first environment vector and the second environment vector corresponding to the same environment data matrix can be subjected to vector parameter mixing processing to form a mixed environment vector corresponding to each of the A environment data matrices. The first environment vector combination includes the first environment vector corresponding to each of the A environment data matrices, and the second environment vector combination includes the second environment vector corresponding to each of the A environment data matrices. For example, the first environment vector and the second environment vector corresponding to the first environment data matrix can be subjected to vector parameter mixing processing to form a mixed environment vector corresponding to the first environment data matrix. Similarly, the first environment vector and the second environment vector corresponding to the second environment data matrix can be subjected to vector parameter mixing processing to form a mixed environment vector corresponding to the second environment data matrix.

[0120] Step S126b: Determine the combination of mixed environment vectors based on the mixed environment vectors corresponding to each of the A environmental data matrices.

[0121] In this embodiment of the invention, after the mixed environment vector is formed, the mixed environment vector combination can be determined based on the mixed environment vectors corresponding to each of the A environment data matrices. For example, the mixed environment vectors can be arranged and combined based on the order of the various environment data matrices to form a mixed environment vector combination.

[0122] Step S126c: Based on the target filtering vector combination and the mixed environment vector combination, determine the global vector of environmental data.

[0123] In this embodiment of the invention, after forming the hybrid environment vector combination, a global environment data vector can be determined based on the target filtering vector combination and the hybrid environment vector combination. That is, the target filtering vector combination and the hybrid environment vector combination can be further semantically fused to obtain the global environment data vector. For example, the vectors in the target filtering vector combination can be concatenated to obtain a first concatenated vector, and the vectors in the hybrid environment vector combination can be concatenated to obtain a second concatenated vector. Then, if the first and second concatenated vectors have the same size, they can be added to obtain the target vector. Alternatively, if the first and second concatenated vectors have different sizes, they can be fully connected to adjust their sizes, resulting in a first and second fully connected vector with the same size. Then, the first and second fully connected vectors are added to obtain the target vector. Finally, the target vector can be further convolutionally mined and pooled to obtain the global environment data vector (in other embodiments, the target vector can also be directly used as the global environment data vector).

[0124] Optionally, the specific method of mixing vector parameters in step S126a is not limited. For example, in one feasible implementation, in order to fully fuse the semantic information of the two environmental vectors obtained by the first semantic mining and the second semantic mining respectively, step S126a may further include steps a1, a2, a3, a4 and a5, as follows (wherein, the A environmental data matrices include the first environmental data matrix, which can be any environmental data matrix).

[0125] Step a1: Determine the first environment vector corresponding to the first environment data matrix from the first combination of environment vectors, and determine the second environment vector corresponding to the first environment data matrix from the second combination of environment vectors.

[0126] In this embodiment of the invention, a first environment vector corresponding to the first environment data matrix can be determined from the first combination of environment vectors, and a second environment vector corresponding to the first environment data matrix can be determined from the second combination of environment vectors.

[0127] Step a2: Determine the first weight parameter and the second weight parameter.

[0128] In this embodiment of the invention, a first weight parameter and a second weight parameter can be determined. The first weight parameter and the second weight parameter can be formed during the corresponding learning and training process.

[0129] Step a3: Based on the first weight parameter, adjust the parameters of the first environment vector corresponding to the first environment data matrix to form a first environment adjustment vector.

[0130] In this embodiment of the invention, after determining the first weight parameter, the first environment vector corresponding to the first environment data matrix can be adjusted based on the first weight parameter to form a first environment adjustment vector. For example, the first weight parameter and the first environment vector can be multiplied to achieve weighting, thereby obtaining the first environment adjustment vector.

[0131] Step a4: Based on the second weight parameter, adjust the parameters of the second environment vector corresponding to the first environment data matrix to form a second environment adjustment vector.

[0132] In this embodiment of the invention, after determining the second weight parameter, the second environment vector corresponding to the first environment data matrix can be adjusted based on the second weight parameter to form a second environment adjustment vector. For example, the second weight parameter and the second environment vector can be multiplied to achieve weighting, thereby obtaining the second environment adjustment vector.

[0133] Step a5: Mix the vector parameters of the first environment adjustment vector and the second environment adjustment vector to form a mixed environment vector corresponding to the first environment data matrix.

[0134] In this embodiment of the invention, after forming the first environment adjustment vector and the second environment adjustment vector, the first environment adjustment vector and the second environment adjustment vector can be mixed by vector parameter mixing to form a mixed environment vector corresponding to the first environment data matrix, that is, the semantic information of the two environment adjustment vectors is further fused.

[0135] Optionally, in step a2 above, the specific method of determining the first weight parameter and the second weight parameter is not limited. For example, in one feasible implementation, step a2 above may further include the following:

[0136] The first step is to determine a random parameter in the target value range (e.g., 0-1), and to determine the difference between the upper limit of the target value range (e.g., 1) and the random parameter. Also, to obtain the target tuning parameter formed during the training process (i.e., formed during the corresponding training process; it should be noted that the region of the target tuning parameter can be greater than or equal to 0).

[0137] The second step involves performing exponential operations on the random parameter and the difference based on the target tuning parameters to obtain a first calculated value and a second calculated value (for example, the first calculated value can be obtained by performing exponential operations on the random parameter based on the reciprocal of the target tuning parameters, or by performing exponential operations on the difference based on the reciprocal of the target tuning parameters). Based on the first calculated value and the second calculated value, a first weight parameter is determined (for example, the sum of the first calculated value and the second calculated value can be calculated, and the ratio of the first calculated value to the sum can be calculated, or the ratio of twice the first calculated value to the sum can be calculated, thereby obtaining the first weight parameter).

[0138] The third step is to determine the second weight parameter based on the first weight parameter, wherein the first weight parameter and the second weight parameter have a negative correlation, for example, the sum of the first weight parameter and the second weight parameter is equal to 1.

[0139] Optionally, in step a5 above, the specific method of mixing the vector parameters of the first environment adjustment vector and the second environment adjustment vector is not limited. For example, in one feasible implementation, step a5 above may further include the following:

[0140] The first step involves determining a first local vector within the first environment adjustment vector, and then determining a second local vector within the second environment adjustment vector based on the distribution position of the first local vector within the first environment adjustment vector. The distribution position of the second local vector within the second environment adjustment vector is completely different from the distribution position of the first local vector within the first environment adjustment vector. For example, the first local vector can be determined by the distribution position of parameters equal to 1 in a screening matrix, and the second local vector can be determined by the distribution position of parameters equal to 0. The size of this screening matrix is ​​equal to the size of the first environment adjustment vector, and this screening matrix is ​​formed during the learning and training process as a parameter of the corresponding neural network model.

[0141] The second step is to concatenate the first local vector and the second local vector to form a hybrid environment vector corresponding to the first environment data matrix. The concatenation method is not limited. For example, the second local vector can be concatenated after the first local vector, or the concatenation can be performed according to the distribution position of the first local vector in the first environment adjustment vector and the distribution position of the second local vector in the second environment adjustment vector.

[0142] In the second part, regarding step S130, it should be noted that the specific method of anomaly identification based on the global vector of the environmental data is not limited and can be configured according to actual needs.

[0143] For example, in one feasible implementation, the global vector of environmental data can be fully connected to form a fully connected vector of environmental data (size 1*1). Then, an identity mapping or linear mapping can be performed on the fully connected vector of environmental data to obtain a parameter characterizing the degree of anomaly. If the parameter is greater than a threshold, it is determined that an anomaly exists; if the parameter is not greater than the threshold, it is determined that no anomaly exists.

[0144] For example, in another feasible implementation, the global vector of environmental data can be fully connected to form a fully connected environmental data vector (size 1*2). Then, based on a target classification function (such as softmax), the fully connected environmental data vector is activated to form a target probability distribution, wherein the target probability distribution includes the probability that the target environmental monitoring data does not have anomalies and the probability that it has anomalies. Finally, based on the target probability distribution, the environmental anomaly analysis result of the target environmental monitoring data can be determined. For example, if the probability that there are no anomalies is greater than the probability that there are anomalies, then there are no anomalies; if the probability that there are no anomalies is not greater than the probability that there are anomalies, then there are anomalies.

[0145] In summary, the present invention provides a blockchain-based method for storing environmental monitoring data. First, the target environmental monitoring data is matrixed to form an environmental data matrix sequence. Second, multi-angle semantic mining is performed on the environmental data matrix sequence to form multiple semantic mining vectors. Based on these multiple semantic mining vectors, a global environmental data vector is formed. Then, anomaly identification is performed based on the global environmental data vector to generate an environmental anomaly analysis result. On the one hand, when the environmental anomaly analysis result indicates the presence of anomalies, the encrypted environmental monitoring data and the environmental anomaly analysis result are stored in the target blockchain. On the other hand, when the environmental anomaly analysis result indicates the absence of anomalies, the target environmental monitoring data and the environmental anomaly analysis result are also stored in the target blockchain. Based on the above method, different storage schemes are used depending on whether anomalies are present, which can balance data privacy and accessibility. In addition, since the basis for anomaly identification (i.e., the global vector of environmental data) is formed by matrix processing and multi-angle semantic mining of the target environmental monitoring data, the semantic representation accuracy of the identification basis will be high (multi-angle semantic mining after matrix processing can facilitate the discovery of the correlation between environmental data from different environmental monitoring devices). Thus, the reliability of environmental anomaly analysis results can be guaranteed, thereby making the corresponding storage more reliable and improving the problem of effectively balancing data privacy and accessibility in existing technologies.

[0146] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for storing environmental monitoring data based on blockchain, characterized in that, include: The target environmental monitoring data is matrixed to form an environmental data matrix sequence, wherein each environmental data matrix in the environmental data matrix sequence includes environmental data obtained by multiple environmental monitoring devices within the same time segment; In the environmental data matrix sequence, A environmental data matrices are identified; the A environmental data matrices undergo first semantic mining to form a first environmental vector combination; the A environmental data matrices undergo second semantic mining to form a second environmental vector combination; candidate screening indicators are identified, and A screening parameters with sequential relationships among the candidate screening indicators are assigned values ​​to form a first screening parameter distribution, and a second screening parameter distribution is generated based on the first screening parameter distribution, wherein the assignment results include a first value and a second value, the first value is used to reflect the corresponding vector being selected, and the second value is used to reflect the corresponding vector being discarded, and the assignment results of the screening parameters at corresponding coordinates in the first screening parameter distribution and the second screening parameter distribution are different; the first environmental vector combination is screened based on the first screening parameter distribution to form a first screening vector combination, and the second environmental vector combination is screened based on the second screening parameter distribution to form a second screening vector combination, and a target screening vector combination is formed based on the first screening vector combination and the second screening vector combination; the target screening vector combination, the first environmental vector combination, and the second environmental vector combination are fused to determine the global environmental data vector; Anomaly identification is performed based on the global vector of the environmental data to form an environmental anomaly analysis result of the target environmental monitoring data. When the environmental anomaly analysis results indicate that the target environmental monitoring data is abnormal, the target environmental monitoring data is encrypted to form encrypted environmental monitoring data, and the encrypted environmental monitoring data and the environmental anomaly analysis results are stored in the target blockchain. When the environmental anomaly analysis results indicate that the target environmental monitoring data does not exhibit any anomalies, the target environmental monitoring data and the environmental anomaly analysis results are stored in the target blockchain.

2. The method for storing environmental monitoring data based on blockchain as described in claim 1, characterized in that, The step of performing first semantic mining on the A environmental data matrices to form a first environmental vector combination includes: Focus mining is performed on any one of the A environmental data matrices to form a first environmental focus vector corresponding to the environmental data matrix. The first environmental focus vector focuses on reflecting the environmental semantic information of the first focused part in the environmental data matrix. Based on the first environmental focus vector, the environmental data matrix is ​​focused and transferred to form a transferred environmental data matrix corresponding to the environmental data matrix; The transfer environment data matrix is ​​subjected to focused mining to form a second environment focused vector corresponding to the transfer environment data matrix. The second environment focused vector focuses on reflecting the environmental semantic information of the second focused part in the transfer environment data matrix. The second focused part is different from the first focused part. The first environment focus vector is deeply mined to form a first environment depth vector, and the second environment focus vector is deeply mined to form a second environment depth vector; The first environment depth vector and the second environment depth vector are merged to form a first environment vector, and the first environment vectors corresponding to the A environment data matrices are combined to form a first environment vector combination.

3. The method for storing environmental monitoring data based on blockchain as described in claim 2, characterized in that, The step of focusing and mining any one of the A environmental data matrices to form a first environmental focus vector corresponding to the environmental data matrix includes: Perform a first convolution operation and a second convolution operation on any one of the A environmental data matrices to form a first environmental convolution vector and a second environmental convolution vector, wherein the first environmental convolution vector and the second environmental convolution vector have the same size. Based on one of the first environment convolution vector and the second environment convolution vector, cross-attention processing or gating adjustment is performed on the other convolution vector to form the first environment focus vector corresponding to the environment data matrix.

4. The method for storing environmental monitoring data based on blockchain as described in claim 2, characterized in that, The step of focusing and shifting the environmental data matrix based on the first environmental focus vector to form a shifted environmental data matrix corresponding to the environmental data matrix includes: Based on the first environmental focus vector, a first focus parameter distribution corresponding to the environmental data matrix is ​​determined, and based on the first focus parameter distribution, a second focus parameter distribution is determined, wherein the first focus parameter distribution is used to reflect the distribution position of the first focus portion, and the second focus parameter distribution is used to reflect the distribution position of the second focus portion. Based on the second focusing parameter distribution and the environmental data matrix, the transfer environmental data matrix is ​​determined, wherein the first value in the second focusing parameter distribution is used to reflect that the environmental data at the corresponding distribution location belongs to the second focusing part, and the second value in the second focusing parameter distribution is used to reflect that the environmental data at the corresponding distribution location does not belong to the second focusing part.

5. The method for storing environmental monitoring data based on blockchain as described in claim 4, characterized in that, The steps of determining the first focusing parameter distribution corresponding to the environmental data matrix based on the first environmental focusing vector, and determining the second focusing parameter distribution based on the first focusing parameter distribution, include: Without changing the vector size, the first environment focusing vector is linearly mapped to form a first environment mapping vector, and the first environment mapping vector is nonlinearly activated to form a first environment activation vector. The mean value of each vector parameter in the first environment activation vector is calculated to obtain the mean value of the vector parameters. Based on the mean value of the vector parameters, the first environment activation vector is binarized and mapped to form the first focusing parameter distribution corresponding to the environment data matrix. The difference between the first value and each parameter in the first focusing parameter distribution is calculated to form the second focusing parameter distribution.

6. The method for storing environmental monitoring data based on blockchain as described in claim 2, characterized in that, The first environment vector combination includes the first environment vector corresponding to each of the A environment data matrices, and the second environment vector combination includes the second environment vector corresponding to each of the A environment data matrices; The step of determining the global vector of environmental data by fusing the target filtering vector combination, the first environmental vector combination, and the second environmental vector combination includes: In the first environment vector combination and the second environment vector combination, the first environment vector and the second environment vector corresponding to the same environment data matrix are respectively subjected to vector parameter mixing processing to form the mixed environment vectors corresponding to the A environment data matrices. Based on the hybrid environment vectors corresponding to each of the A environmental data matrices, a hybrid environment vector combination is determined; Based on the target filtering vector combination and the mixed environment vector combination, a global vector of environmental data is determined.

7. The method for storing environmental monitoring data based on blockchain as described in claim 6, characterized in that, The A environmental data matrices include the first environmental data matrix; The step of mixing the vector parameters of the first environment vector and the second environment vector corresponding to the same environment data matrix in the first environment vector combination and the second environment vector combination to form the mixed environment vector corresponding to each of the A environment data matrices includes: The first environment vector corresponding to the first environment data matrix is ​​determined from the first combination of environment vectors, and the second environment vector corresponding to the first environment data matrix is ​​determined from the second combination of environment vectors. Determine the first weight parameter and the second weight parameter; Based on the first weight parameter, the parameters of the first environment vector corresponding to the first environment data matrix are adjusted to form the first environment adjustment vector; Based on the second weight parameter, the parameters of the second environment vector corresponding to the first environment data matrix are adjusted to form the second environment adjustment vector; The first environment adjustment vector and the second environment adjustment vector are mixed by vector parameter processing to form a mixed environment vector corresponding to the first environment data matrix.

8. The method for storing environmental monitoring data based on blockchain as described in claim 7, characterized in that, The step of mixing the first environment adjustment vector and the second environment adjustment vector to form a mixed environment vector corresponding to the first environment data matrix includes: A first local vector is determined in the first environment adjustment vector, and a second local vector is determined in the second environment adjustment vector based on the distribution position of the first local vector in the first environment adjustment vector, wherein the distribution position of the second local vector in the second environment adjustment vector is completely different from the distribution position of the first local vector in the first environment adjustment vector. The first local vector and the second local vector are concatenated to form the hybrid environment vector corresponding to the first environment data matrix.

9. The method for storing environmental monitoring data based on blockchain as described in any one of claims 1-8, characterized in that, The step of identifying anomalies based on the global vector of the environmental data to form the environmental anomaly analysis result of the target environmental monitoring data includes: The global vector of environmental data is processed by full connection to form a fully connected vector of environmental data; Based on the target classification function, the fully connected vector of the environmental data is activated to form a target probability distribution, wherein the target probability distribution includes the probability that the target environmental monitoring data does not have anomalies and the probability that it has anomalies; Based on the target probability distribution, the environmental anomaly analysis results of the target environmental monitoring data are determined.

Images