A water quality monitoring method based on blockchain-enabled pollutant fingerprint collaborative authentication

Abstract

The present invention discloses a water quality monitoring method based on blockchain-enabled pollutant fingerprint collaborative authentication, which relates to the field of water pollution monitoring. The present invention includes the following steps: collecting fingerprint data of water bodies; preprocessing the fingerprint data into feature data; performing deviation judgment on the hash value of the fingerprint data and the previous record, and performing calibration if the deviation exceeds a threshold, otherwise, writing the feature data into a hash tree; loading the feature data into a pre-trained migration model, with the input data dimension corresponding to the dimension of the feature data, and the migration model pre-trained corresponding to the target water body type, and the migration model outputting a pollution source probability matrix and a confidence score; the present invention ensures the immutability of data through blockchain, and uses smart contracts to realize automated quality control, thereby achieving the effect of "one-time calibration, full domain applicability" in complex water body monitoring.

Description

Technical Field

[0001] The present invention relates to the field of water pollution monitoring, and in particular to a water quality monitoring method based on blockchain-enabled collaborative authentication of pollutant fingerprints. Background Art

[0002] For complex water bodies, these may include rivers or industrial wastewater with high turbidity, multiple ions, and rich organic matter. Secondly, there are combinations of "multiple pollutants," commonly including heavy metals + organic matter, nitrogen and phosphorus nutrients + pesticides, etc. Finally, there are "mutual interference" mechanisms, such as electrode contamination, spectral overlap, and biotoxicity superposition effects.

[0003] The convergence of multiple technologies suggests that spectroscopic-electrochemical instruments can simultaneously circumvent turbidity interference and electrode passivation. However, physically, there is a spatial conflict between the optical path and the circuit. When the laser penetrates the electrolytic cell to illuminate the electrode surface, it is necessary to ensure both the light window transmittance and the electrode active area. Bubbles generated during high-frequency pulse cleaning of the electrode directly distort the spectral signal. Pulse cleaning can alleviate passivation but exacerbates signal fluctuations. Turbidity compensation algorithms sacrifice resolution.

[0004] Therefore, there is an urgent need to improve the applicability and reliability of monitoring technology in complex water environments; to avoid mutual interference between pollutants, it is necessary to comprehensively use a variety of technical strategies and management methods. Summary of the Invention

[0005] The present invention provides a water quality monitoring method based on blockchain-enabled pollutant fingerprint collaborative authentication to solve the problems of the existing technology.

[0006] In a first aspect, the present invention provides a water quality monitoring method based on blockchain-enabled pollutant fingerprint collaborative authentication, comprising the following steps:

[0007] Collect fingerprint data of water bodies;

[0008] Preprocessing fingerprint data into feature data;

[0009] The hash value of the fingerprint data is judged to be different from the previous record. If the deviation exceeds the threshold, calibration is performed. Otherwise, the feature data is written into the hash tree.

[0010] Load feature data into the pre-trained migration model. The input data dimension corresponds to the dimension of the feature data. The migration model is pre-trained for the target water body type. The migration model outputs the pollution source probability matrix and confidence score.

[0011] Among them, the pollution source probability matrix elements represent the probability that the pollution source contributes pollutants.

[0012] Furthermore, the fingerprint data collected of the water body includes: spectral fingerprint data, electrochemical time series data, and biological response fingerprint data.

[0013] Furthermore, the fingerprint data is preprocessed into feature data, including: Fourier transforming the spectral fingerprint data into a 128-dimensional feature vector, implementing wavelet noise reduction encoding on the electrochemical fingerprint data into a 96-dimensional feature matrix, and converting the biological response fingerprint data into a DNA barcode through PCR amplification technology; wherein, the electrochemical time series data is generated into electrochemical fingerprint data after feature extraction.

[0014] Furthermore, the hash value of the fingerprint data is judged to be different from the previous record, and calibration is performed if the deviation exceeds a threshold; otherwise, the feature data is written into the hash tree, specifically including:

[0015] If the deviation between the SHA-3 hash value of the fingerprint data and the previous record exceeds the threshold, the calibration instruction is executed. If the deviation is within the threshold, the feature vector of the fingerprint data is written into the hash tree and a timestamp is generated for storage.

[0016] Use IPFS protocol to store original fingerprint data;

[0017] The preceding data are the benchmark fingerprint data of the same monitoring point in the last calibration period, and the benchmark fingerprint data is the same as the corresponding preceding record in terms of the temporal stability of water pollutants.

[0018] Furthermore, the hash value of the fingerprint data is subjected to deviation discrimination with respect to the preceding record. If the deviation exceeds a threshold, calibration is performed. Otherwise, the feature data is written into the hash tree. The calibration includes:

[0019] When the turbidity detected is greater than 200 NTU, the following steps will be automatically executed:

[0020] S1, switch the sensing mode to anti-interference pulse excitation;

[0021] The technical effect is to reduce the thickness of the diffusion layer and inhibit the adsorption of suspended matter;

[0022] S2, call three adjacent monitoring nodes for cross-validation;

[0023] Cross-validation mechanism: Neighboring nodes will synchronously detect the same water body, and calibration will be triggered when the correlation coefficient between the main node data and the data of ≥2 neighboring nodes is lower than 0.85.

[0024] The cross-validation process includes:

[0025] S21. Broadcast the verification request to neighboring nodes within a geographical radius of 1km;

[0026] S22, the receiving node synchronously collects data and returns a hash value;

[0027] S23. Calculate the consistency index. If the consistency index κ ≥ 0.67, accept the original data.

[0028] S3. Send ERC-20 based on the cross-validation results.

[0029] The essence of ERC-20 is a maintenance incentive mechanism. After calibration, it can be exchanged for cash or priority use of the equipment, that is, exchanged for equipment maintenance services or discounts on monitoring service fees;

[0030] Furthermore, the migration model setting includes:

[0031] Deploy edge computing nodes around monitoring points, which run lightweight CNN models to classify pollutant types in real time;

[0032] The cloud server periodically aggregates the gradients of each node to update the global model weights.

[0033] Further, including: Training settings for migration models:

[0034] Preset value, add conditional normalization to the last layer of the migration model, and input the learnable water type embedding vector into the normalization process;

[0035] Pre-training: The pre-training objective function is set as the pollutant classification cross entropy loss, and the multi-basin mixed data training set is used for training;

[0036] Fine-tune the model, collect multiple sets of labeled samples of the target water area, set the fine-tuning objective function to contrast loss with domain adaptation, and perform training fine-tuning.

[0037] Furthermore, biological data serves as an independent chain of evidence to verify with chemical data, namely DNA barcodes and electrochemical fingerprint data, spectral fingerprint data. The sensor directly uploads the DNA barcode to the blockchain through the IPFS protocol. When the migration model runs the biological toxicity model, the detected toxicity index is sent to the blockchain. When it is found that the biological toxicity index exceeds the standard, the regulatory agency retrieves the DNA barcode evidence package from the blockchain.

[0038] Furthermore, the loss functions used in training the migration model include:

[0039] Pollutant classification cross entropy loss function, including:

[0040] Multi-label design corresponding to pollutant types;

[0041] The weights of different categories are positively correlated with the toxicity level of pollutants;

[0042] The output layer uses sigmoid to achieve multi-label prediction;

[0043] Contrastive loss function with domain adaptation, including:

[0044] Positive sample pair: feature vector of the same pollutant in different water areas;

[0045] Negative sample pairs: feature vectors of different pollutants in the same water area;

[0046] A distance metric in the feature space is used to adjust parameters and cluster similar pollutants across domains.

[0047] The feature embedding learned by the contrastive loss function with domain adaptation is used as the feature input of the cross entropy loss function for pollutant classification and jointly optimized through gradient coupling.

[0048] This invention ensures data immutability through blockchain and uses smart contracts to achieve automated quality control, achieving the effect of "one-time calibration, full application" in complex water body monitoring.

[0049] The migration model based on federated learning in the present invention breaks through the limitations of environmental differences and provides monitoring adaptation accuracy in different waters. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of the present invention, and do not constitute a limitation of the embodiments of the present invention. In the drawings:

[0051] Figure 1 A flow chart of a water quality monitoring method based on blockchain-enabled collaborative authentication of pollutant fingerprints provided as an exemplary embodiment of the present invention.

[0052] Figure 2 A flowchart of a method provided in accordance with an exemplary embodiment of the present invention, in which a water quality monitoring method based on blockchain-enabled pollutant fingerprint collaborative authentication is automatically executed when a turbidity greater than 200 NTU is detected. DETAILED DESCRIPTION

[0053] Exemplary embodiments will be described in detail herein, examples of which are illustrated in the accompanying drawings. In the following description, when referring to the drawings, like numbers in different figures represent like or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all possible embodiments consistent with the present invention. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present invention, as detailed in the appended claims.

[0054] This invention considers covering various water environments, from plateau lakes to industrial estuaries. First, the technical concept is as follows: the formats of spectral and electrochemical raw data collected by different equipment vary greatly, and there are different systematic errors. Therefore, this invention considers designing a unified digital fingerprint encoding specification. To adapt to subsequent processing rules, the multimodal data is fused into a 128-dimensional feature vector, which not only retains key information but also facilitates storage.

[0055] Consider using a consortium chain, which is superior to a public chain. Environmental protection agencies, research institutions, and enterprises will be used as nodes to ensure data authority and improve decentralization efficiency. For abnormal data, it can be set to automatically trigger a confidence-weighted arbitration algorithm.

[0056] By training the knowledge graph of pollutants offline, the recognition accuracy can be improved, and a lightweight model is used online to process real-time diagnostic data at the edge.

[0057] The specific application scenario of this invention is water pollution monitoring in complex water bodies. This invention combines the blockchain-based evidence storage mechanism of pollutant fingerprints, dynamic calibration driven by smart contracts, and the cross-water migration capability of large-scale models to jointly solve the problem of complex water body monitoring.

[0058] The present invention provides a water quality monitoring method based on blockchain-enabled pollutant fingerprint collaborative authentication, which aims to solve the above technical problems of the prior art.

[0059] The following describes in detail the technical solution of the present invention and how the technical solution of the present invention solves the above-mentioned technical problems using specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The following embodiments of the present invention are described in conjunction with the accompanying drawings.

[0060] Example 1:

[0061] This embodiment provides a water quality monitoring method based on blockchain-enabled pollutant fingerprint collaborative authentication, such as Figure 1 、 Figure 2 The following steps are shown:

[0062] A1. Collect fingerprint data of water bodies;

[0063] Including: spectral fingerprint data, electrochemical time series data, biological response fingerprint data;

[0064] A2. Preprocessing fingerprint data into feature data;

[0065] The spectral fingerprint data was Fourier transformed into a 128-dimensional feature vector, and the electrochemical fingerprint data was encoded into a 96-dimensional feature matrix using wavelet noise reduction. The bioresponse fingerprint data was then converted into a DNA barcode using PCR amplification. The electrochemical time series data was then subjected to feature extraction to generate the electrochemical fingerprint data. Specifically, the electrochemical time series data was subjected to wavelet noise reduction to extract the peak current (i_p), peak potential (E_p), and peak separation (ΔE_p) parameters, forming the electrochemical feature matrix, i.e., the electrochemical fingerprint data.

[0066] The choice of feature dimension, 128 or 96, is based on a balance between actual sensor resolution and information entropy. UV-VIS spectrometers typically output 2048 raw data points. Based on the standard resolution of UV-VIS spectrometers (2nm interval, 301 raw data points in the 200-800nm ​​range), retaining the first 128 low-frequency components after Fourier transform compression can cover over 95% of the feature information. Cyclic voltammetry data collected by an electrochemical workstation typically has 1024 sampling points. After threshold filtering using a wavelet transform, 96 dimensions optimally preserve the redox characteristic peaks. 32 characteristic peaks x 3 parameters = 96 dimensions, and exceeding 96 dimensions increases the risk of model overfitting (cross-validation accuracy decreases by ≥3%).

[0067] A3. Determine the deviation between the hash value of the fingerprint data and the previous record. If the deviation exceeds a threshold, perform calibration. Otherwise, write the feature data into the hash tree.

[0068] Specifically include:

[0069] If the deviation between the SHA-3 hash value of the fingerprint data and the previous record exceeds the threshold, the calibration instruction is executed. If the deviation is within the threshold, the feature vector of the fingerprint data is written into the hash tree and a timestamp is generated for storage.

[0070] Use IPFS protocol to store original fingerprint data;

[0071] The preceding data are the benchmark fingerprint data of the same monitoring point in the last calibration period, and the benchmark fingerprint data is the same as the corresponding preceding record in terms of the temporal stability of water pollutants.

[0072] Previous record: the median of the fingerprint hash value of the sliding window in the previous 7 days at the same monitoring point;

[0073] That is, relative to the length of the calibration period, water pollutants can be regarded as static in the two adjacent periods. At the same time, the regularization term of the objective function under the assumption of temporal stability of water pollutants constitutes the core of transfer learning. This term constrains the parameters of the new water model to be too far from the basic model to avoid overfitting.

[0074] Calibration includes:

[0075] The dynamic calibration sub-contract is deployed on the Ethereum virtual machine and set in the smart contract unit. The measurement source is the 90° nephelometric turbidimeter (ISO 7027 standard) integrated into the sensor module. The trigger logic: the contract is activated when the turbidity value of the sample is >200 NTU for five consecutive times;

[0076] When the turbidity detected is greater than 200 NTU, the following steps will be automatically executed:

[0077] S1, switch the sensing mode to anti-interference pulse excitation;

[0078] Switch the potentiostatic mode to:

[0079] Pulse width: 50ms;

[0080] Interval time: 20ms;

[0081] Potential step: +0.2 V → -0.3 V (vs. Ag / AgCl);

[0082] The technical effect is to reduce the thickness of the diffusion layer and inhibit the adsorption of suspended matter;

[0083] S2, call three adjacent monitoring nodes for cross-validation;

[0084] Cross-validation mechanism: Neighboring nodes will synchronously detect the same water body, and calibration will be triggered when the correlation coefficient between the main node data and the data of ≥2 neighboring nodes is lower than 0.85.

[0085] The cross-validation process includes:

[0086] S21. Broadcast the verification request to neighboring nodes within a geographical radius of 1km;

[0087] S22, the receiving node synchronously collects data and returns a hash value;

[0088] S23. Calculate the consistency index. If the consistency index κ ≥ 0.67, accept the original data.

[0089] S3. Send ERC-20 based on the cross-validation results.

[0090] The essence of ERC-20 is a maintenance incentive mechanism. After calibration, it can be exchanged for cash or priority use of equipment, that is, exchanged for equipment maintenance services or discounts on monitoring service fees.

[0091] A4. Load feature data into the pre-trained migration model. The input data dimensions correspond to the feature data dimensions. The migration model is pre-trained for the target water body type. The migration model outputs a pollution source probability matrix and confidence score.

[0092] The transfer learning model adapts the domain adaptation layer according to the target water body type and dynamically adjusts the model parameters.

[0093] The objective function of the transfer model is set based on the federated learning loss and the domain adaptation regularization term.

[0094] Among them, the pollution source probability matrix elements represent the probability that the pollution source contributes pollutants.

[0095] The migration model settings include:

[0096] Deploy edge computing nodes around monitoring points, which run lightweight CNN models to classify pollutant types in real time;

[0097] The cloud server periodically aggregates the gradients of each node to update the global model weights.

[0098] For the training settings of the migration model:

[0099] Preset value, add conditional normalization to the last layer of the migration model, and input the learnable water type embedding vector into the normalization process;

[0100] Pre-training: The pre-training objective function is set as the pollutant classification cross entropy loss, and the multi-basin mixed data training set is used for training;

[0101] Fine-tune the model, collect multiple sets of labeled samples of the target water area, set the fine-tuning objective function to contrast loss with domain adaptation, and perform training fine-tuning.

[0102] The loss functions used in training the migration model include:

[0103] Pollutant classification cross entropy loss function, including:

[0104] Multi-label design corresponding to pollutant types;

[0105] The weights of different categories are positively correlated with the toxicity level of pollutants;

[0106] The output layer uses sigmoid to achieve multi-label prediction;

[0107] Contrastive loss function with domain adaptation, including:

[0108] Positive sample pair: feature vector of the same pollutant in different water areas;

[0109] Negative sample pairs: feature vectors of different pollutants in the same water area;

[0110] A distance metric in the feature space is used to adjust parameters and cluster similar pollutants across domains.

[0111] The feature embedding learned by the contrastive loss function with domain adaptation is used as the feature input of the cross entropy loss function for pollutant classification and jointly optimized through gradient coupling.

[0112] Preferably, when biological data is used as an independent chain of evidence, it can be cross-verified with chemical data, that is, DNA barcodes and electrochemical fingerprint data, spectral fingerprint data. The sensor directly uploads the DNA barcode to the blockchain through the IPFS protocol. When the migration model runs the biological toxicity model, the detected toxicity index is sent to the blockchain. When it is found that the biological toxicity index exceeds the standard, the regulatory agency retrieves the DNA barcode evidence package from the blockchain.

[0113] Example 2:

[0114] Based on Example 1, a specific example of step A3 is to execute a calibration instruction if the deviation between the SHA-3 hash value of the fingerprint data and the previous record exceeds a threshold, that is, when the smart contract detects that |H_{new} - H_{baseline}|>5%: the threshold is 5%.

[0115] Sending a calibration command to the device and updating the new reference hash value to the hash standard response curve;

[0116] Input data: digital fingerprint after feature extraction (128-dimensional vector + 96-dimensional matrix);

[0117] Hash algorithm: SHA3-256;

[0118] Tree structure construction:

[0119] def build_merkle_tree(hashes):

[0120] if len(hashes) == 1:

[0121] return hashes[0]

[0122] new_level = []

[0123] for i in range(0, len(hashes), 2):

[0124] left = hashes[i]

[0125] right = hashes[i+1]if i+1 <len(hashes) else left

[0126] new_level.append(sha3(left + right)) # Node merge hash

[0127] return build_merkle_tree(new_level)

[0128] Among them, the content on the chain: only the root hash value (fixed 256 bits) and the timestamp;

[0129] Achieve double tamper-proofing: Data level: Modify the original file → CID changes → does not match the on-chain record;

[0130] Structural level: Modify any feature → Hash root hash changes → Blockchain consensus rejection;

[0131] To verify data integrity, you only need to provide the hash path, without downloading the entire original data (only about 1KB of proof data is required).

[0132] In the several embodiments provided herein, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the module division is merely a logical functional division. In actual implementation, other division methods may be used, such as combining or integrating multiple modules or components into another system, or omitting or not implementing certain features.

[0133] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules may be selected to achieve the purpose of the present embodiment according to actual needs.

[0134] In addition, the functional modules in various embodiments of the present invention may be integrated into a single processing module, each module may exist physically separately, or two or more modules may be integrated into a single module. The aforementioned integrated modules may be implemented in the form of hardware or hardware plus software functional modules.

[0135] Those skilled in the art will appreciate that the embodiments of the present invention may be provided as methods or systems. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware.

[0136] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0137] The above are merely embodiments of the present invention and are not intended to limit the present invention. It will be apparent to those skilled in the art that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention are intended to be included within the scope of the claims of the present invention.

[0138] Those skilled in the art will readily recognize other embodiments of the present invention after considering the invention disclosed herein in the specification and examples. The present invention is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the invention being indicated by the following claims.

[0139] It should be understood that the present invention is not limited to the exact construction described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof, which is limited only by the appended claims.

Claims

1. A water quality monitoring method based on blockchain-enabled pollutant fingerprint collaborative authentication, characterized in that: The steps include: Collect fingerprint data of water bodies; Preprocessing fingerprint data into feature data; The hash value of the fingerprint data is judged to be different from the previous record. If the deviation exceeds the threshold, calibration is performed. Otherwise, the feature data is written into the hash tree. Specifically, the following steps are performed: If the deviation between the SHA-3 hash value of the fingerprint data and the previous record exceeds the threshold, the calibration instruction is executed. If the deviation is within the threshold, the feature vector of the fingerprint data is written into the hash tree and a timestamp is generated for storage. Use IPFS protocol to store original fingerprint data; The preceding record is the baseline fingerprint data of the same monitoring point in the last calibration period, and the time stability of the baseline fingerprint data depending on the water pollutants is the same as that of the corresponding preceding record; Load feature data into the pre-trained migration model. The input data dimension corresponds to the dimension of the feature data. The migration model is pre-trained for the target water body type. The migration model outputs the pollution source probability matrix and confidence score. Among them, the elements of the pollution source probability matrix represent the probability that the pollution source contributes pollutants.

2. A water quality monitoring method based on blockchain-enabled pollutant fingerprint collaborative authentication according to claim 1, characterized in that: The fingerprint data of the water body collected includes: spectral fingerprint data, electrochemical time series data, and biological response fingerprint data.

3. A water quality monitoring method based on blockchain-enabled pollutant fingerprint collaborative authentication according to claim 2, characterized in that: The fingerprint data is preprocessed into feature data, including: Fourier transforming the spectral fingerprint data into a 128-dimensional feature vector, performing wavelet noise reduction encoding on the electrochemical fingerprint data into a 96-dimensional feature matrix, and converting the biological response fingerprint data into a DNA barcode through PCR amplification technology; wherein, the electrochemical time series data is subjected to feature extraction to generate the electrochemical fingerprint data.

4. A water quality monitoring method based on blockchain-enabled pollutant fingerprint collaborative authentication according to claim 3, characterized in that: The deviation between the hash value of the fingerprint data and the previous record is judged, and calibration is performed if the deviation exceeds a threshold; otherwise, the feature data is written into the hash tree, wherein the calibration includes: When the turbidity detected is greater than 200 NTU, the following steps will be automatically executed: S1, switch the sensing mode to anti-interference pulse excitation; S2, call three adjacent monitoring nodes for cross-validation; S3. Send ERC-20 based on the cross-validation results.

5. A water quality monitoring method based on blockchain-enabled pollutant fingerprint collaborative authentication according to claim 4, characterized in that: in, For migration model settings, include: Deploy edge computing nodes around monitoring points, which run lightweight CNN models to classify pollutant types in real time; The cloud server periodically aggregates the gradients of each node to update the global model weights.

6. A water quality monitoring method based on blockchain-enabled pollutant fingerprint collaborative authentication according to claim 5, characterized in that: include: For the training settings of the migration model: Preset value, add conditional normalization to the last layer of the migration model, and input the learnable water type embedding vector into the normalization process; Pre-training: The pre-training objective function is set as the pollutant classification cross entropy loss, and the multi-basin mixed data training set is used for training; Fine-tune the model, collect multiple sets of labeled samples of the target water area, set the fine-tuning objective function to contrast loss with domain adaptation, and perform training fine-tuning.

7. A water quality monitoring method based on blockchain-enabled pollutant fingerprint collaborative authentication according to claim 3, characterized in that: in, Biological data serves as an independent chain of evidence and is cross-verified with chemical data, namely DNA barcodes, electrochemical fingerprint data, and spectral fingerprint data. The sensor directly uploads the DNA barcode to the blockchain through the IPFS protocol. When the migration model runs the biological toxicity model, the detected toxicity index is sent to the blockchain. When it is found that the biological toxicity index exceeds the standard, the regulatory agency retrieves the DNA barcode evidence package from the blockchain.

8. The water quality monitoring method based on blockchain-enabled pollutant fingerprint collaborative authentication according to claim 6 is characterized in that: The loss functions used in training the migration model include: Pollutant classification cross entropy loss function, including: Multi-label design corresponding to pollutant types; The weights of different categories are positively correlated with the toxicity level of pollutants; The output layer uses sigmoid to achieve multi-label prediction; Contrastive loss function with domain adaptation, including: Positive sample pair: feature vector of the same pollutant in different water areas; Negative sample pairs: feature vectors of different pollutants in the same water area; A distance metric in the feature space is used to adjust parameters and cluster similar pollutants across domains. The feature embedding learned by the contrastive loss function with domain adaptation is used as the feature input of the cross entropy loss function for pollutant classification and jointly optimized through gradient coupling.

Images