A water index analysis method applied to complex water bodies

Abstract

The invention discloses a water body index analysis method applied to complex water bodies, and relates to the field of water body index intelligent detection algorithms. The invention performs denoising and turbidity compensation by adaptively adjusting filter coefficients and thresholds, thereby improving the ability to capture multidimensional nonlinear features under complex water body conditions. In addition, the invention also adopts a nonlinear regression model based on a multidimensional kernel function to predict indicators, and extracts features through multi-level recursion, thereby improving the accuracy of prediction.

Description

Technical Field

[0001] The present invention relates to the field of intelligent detection algorithms for water body indicators, and in particular to a water body indicator analysis method applied to complex water bodies. Background Art

[0002] Complex water bodies typically refer to urban rivers, lakes, and reservoirs. These water bodies are subject to multiple pollution sources, including industrial emissions, agricultural non-point source pollution, and domestic sewage discharge. Environmental characteristics of complex water bodies include turbidity, particulate matter concentration, suspended solids size and distribution, and light scattering properties, which complicate water quality monitoring and analysis. Furthermore, water index analysis is a crucial component of environmental monitoring and water quality management. Traditional water index analysis methods rely on chemical reagents and laboratory equipment. While accurate, these methods are time-consuming and costly. With the advancement of remote sensing and spectral analysis technologies, spectral-based water index analysis methods have become a research hotspot. These methods analyze the absorption characteristics of water at specific wavelengths of light and use established prediction models to quantitatively analyze water quality parameters. Commonly used prediction models include linear models such as single-wavelength methods, multi-wavelength combination methods, and partial least squares (PLS), as well as nonlinear models such as support vector machines (SVMs), neural networks, and decision trees.

[0003] Existing water monitoring technologies often rely on periodic field sampling and laboratory analysis. While accurate, these methods are time-consuming and costly, making them unsuitable for rapid response and large-scale monitoring. Furthermore, traditional monitoring techniques may not fully capture the multidimensional, nonlinear characteristics of complex water bodies, such as changes in turbidity and particulate matter concentration, and how these factors affect the accuracy of spectral data. Existing technologies often struggle to achieve real-time, continuous, and high-frequency monitoring of water quality parameters when dealing with complex water bodies, limiting the ability to rapidly respond to dynamic changes in water pollution. Summary of the Invention

[0004] The present invention proposes a water index analysis method applied to complex water bodies. This method effectively improves the accuracy and efficiency of water index analysis through advanced technologies such as adaptive filtering, multi-scale adaptive wavelet transform, nonlinear kernel function and high-order tensor operation.

[0005] Among them, a water index analysis method applied to complex water bodies includes the following steps:

[0006] S1. Extract spectral data from complex water bodies and preprocess the raw spectral data using adaptive filtering. The adaptive filter adjusts the filter coefficient based on changes in the water environment and outputs preprocessed, smoothed spectral data.

[0007] S2. Decompose the preprocessed data into different frequency components using a multi-scale adaptive wavelet transform and perform denoising using an adaptive threshold.

[0008] S3. The turbidity of the denoised spectral data is compensated using nonlinear kernel functions and high-order tensor operations.

[0009] S4. Extracting features from the compensated spectral data through multi-level recursion and predicting water quality indicators of complex water bodies using a regression model, including total phosphorus, ammonia nitrogen, and total nitrogen;

[0010] Wherein, the step S3 specifically includes the following sub-steps:

[0011] S301. Based on the nonlinear effect of turbidity on the spectrum, the spectral signal is mapped to a high-dimensional space using a kernel method to capture nonlinear features;

[0012] S302. When the characteristics of different wavelengths and time points affect each other, multi-dimensional coupling compensation is performed on the spectral data through a high-order tensor compensation model;

[0013] S303. The nonlinear compensation value output by the kernel method is combined with the tensor compensation term output by the high-order tensor compensation model, a total compensation calculation is performed on the shading degree, and the compensated spectral data is output.

[0014] Furthermore, in step S301, the spectral signal is mapped to a high-dimensional space by a kernel method to capture nonlinear features. The specific process is expressed as follows:

[0015] ;

[0016] Among them, the It represents the turbidity compensation item after compensation, covering the compensation of nonlinear changes of spectral data at different wavelengths and time dimensions. represents the wavelength of the spectral data, Represents the time dimension, that is, the time point of collecting spectral data. Indicates that the wavelength of the spectral data during the turbidity compensation process is The total number of and Indicates the corresponding spectral signal at wavelength and The spectral signal index under Indicates that the wavelength of the spectral data during the turbidity compensation process is The total number of represents the kernel function, represents the spectral data after denoising, represents the pth wavelength point, represents the qth wavelength point, represents the derivative operation, Indicates the adjustment coefficient, used to adjust the wavelength The influence of the second-order derivative at on the compensation results.

[0017] Furthermore, in step S302, performing multi-dimensional coupling compensation on the spectral data by using a high-order tensor compensation model specifically includes the following sub-steps:

[0018] S3021. Model the nonlinear coupling relationship between different wavelengths and time dimensions using a third-order tensor. Express the nonlinear coupling between different wavelengths in complex water bodies using tensor operations, namely:

[0019] ;

[0020] Among them, the represents the turbidity compensation value obtained after calculation by the third-order tensor model. 、 and Respectively represent the wavelength index, the Represents the tensor coefficient of the third-order tensor, reflecting the wavelength 、 and The coupling weight at time t, represents the total number of wavelength dimensions, represents the wavelength of the spectral data, Represents the time dimension, 、 and Represent the i-th wavelength point, j-th wavelength point and k-th wavelength point respectively;

[0021] S3022. Capture the nonlinear variations of spectral signals at different wavelengths and time dimensions through tensor compensation. Further adjust the compensation results based on the adjustment parameters and the influence of the second-order derivatives at different wavelengths and time dimensions, namely:

[0022] ;

[0023] Among them, the Represents the turbidity value after high-order tensor compensation, Represents a scalar parameter used to adjust the weight of the tensor compensation result. represents the derivative operation, Represents the elements of the coefficient matrix, which reflects the coupling strength between the i-th and j-th wavelengths.

[0024] Furthermore, the specific process of step S303 is as follows:

[0025] ;

[0026] Among them, the represents the spectral data after compensation, represents the wavelength of the spectral data, Represents the time dimension, represents the spectral data after denoising, Indicates the turbidity compensation item after compensation. Indicates the turbidity value after high-order tensor compensation.

[0027] Furthermore, the step S4 specifically includes the following sub-steps:

[0028] S401. Perform recursive feature extraction on the pre-processed compensated spectral data, using step-by-step recursion to capture multidimensional nonlinear features under complex water conditions. The step-by-step recursion of multidimensional nonlinear features includes primary feature extraction and secondary feature extraction.

[0029] S402. Predicting indicators based on the extracted features using a nonlinear regression model based on a multidimensional kernel function;

[0030] S403. Integrate the predicted values ​​at all levels calculated by the piecewise nonlinear regression model to generate accurate predictions of water body indicators.

[0031] Furthermore, in step S401, the first-level feature extraction is specifically as follows: based on the first-level features, recursive feature extraction is performed in combination with features related to the total phosphorus index in the spectral data, namely:

[0032] ;

[0033] Among them, the Represents the first-level feature extraction result, which is used to extract spectral features related to the total phosphorus index. represents the wavelength of the spectral data, Represents the time dimension, Indicates the number of features extracted from the first-level feature. Represents the index of the feature, It represents the weight coefficient of the kth feature in the first-level feature extraction, which is used to characterize the importance of this feature in the prediction of total phosphorus index. Represents the basis function of the kth feature in the first-level feature extraction, which is used to describe the specific method of feature extraction. represents the spectral data after compensation, Represents the adjustment parameters related to the total phosphorus index in the spectral data.

[0034] Furthermore, in step S401, the secondary feature extraction is specifically: recursive feature extraction is performed in combination with the features of the ammonia nitrogen or total nitrogen index in the spectral data, that is:

[0035] ;

[0036] Among them, the Represents the secondary feature extraction result, which is used to extract spectral features related to ammonia nitrogen or total nitrogen indicators. represents the wavelength of the spectral data, Represents the time dimension, Indicates that Indicates the number of features extracted from the secondary feature. Represents the weight coefficient of the kth feature in the secondary feature extraction, Represents the basis function of the kth feature in the secondary feature extraction, Represents the first-level feature extraction result, Indicates the adjustment parameters related to ammonia nitrogen or total nitrogen indicators.

[0037] Furthermore, in step S402, the nonlinear regression model based on the multidimensional kernel function is specifically expressed as:

[0038] ;

[0039] Among them, the Represents the predicted value of the i-th water body index, represents the wavelength of the spectral data, Represents the time dimension, Represents the number of features related to the i-th water body index, Indicates the total number of features related to water body indicators, Represents the nonlinear kernel function of the i-th indicator, Represents the feature extraction result of the i-th water body index, Represents adjustment parameters related to multidimensional environmental parameters, Represents the weight coefficient of the j-th feature.

[0040] Furthermore, the multi-dimensional environmental parameters include: turbidity of water, concentration of particulate matter, size and distribution of suspended matter, and light scattering characteristics.

[0041] Furthermore, in step S403, the specific process of integrating the predicted values ​​is as follows:

[0042] ;

[0043] Among them, the Represents the comprehensive water index prediction value, Indicates the total number of water body indicators, Represents the index of water body index, Represents the predicted value of the i-th water body index, represents the wavelength of the spectral data, Represents the time dimension.

[0044] The beneficial effects of the invention are:

[0045] The present invention performs denoising and turbidity compensation by adaptively adjusting the filter coefficient and threshold, thereby improving the ability to capture multidimensional nonlinear features under complex water conditions. In addition, the present invention also adopts a nonlinear regression model based on a multidimensional kernel function to predict indicators and extracts features through multi-level recursion to improve the accuracy of prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 A flow chart of a method for analyzing water body indicators applied to complex water bodies provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings, but the protection scope of the present invention is not limited to the following.

[0048] In order to make the objectives, technical solutions, and advantages of the present invention more clearly understood, the present invention is further described in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only intended to explain the present invention and are not intended to limit the present invention. That is, the embodiments described herein are only some embodiments of the present invention, not all embodiments. Generally, the components of the embodiments of the present invention described and illustrated in the drawings herein may be arranged and designed in various different configurations.

[0049] 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 represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative work shall fall within the scope of protection of the present invention. It should be noted that relational terms such as "first" and "second" are merely used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations.

[0050] Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not preclude the presence of additional identical elements in the process, method, article, or apparatus that includes the element.

[0051] The features and performance of the present invention are further described in detail below with reference to the embodiments.

[0052] Among them, such as Figure 1 , a water body index analysis method applied to complex water bodies, comprising the following steps:

[0053] S1. Extract spectral data from complex water bodies and preprocess the raw spectral data using adaptive filtering. The adaptive filter adjusts the filter coefficient based on changes in the water environment and outputs preprocessed, smoothed spectral data.

[0054] S2. Decompose the preprocessed data into different frequency components using a multi-scale adaptive wavelet transform and perform denoising using an adaptive threshold.

[0055] S3. The turbidity of the denoised spectral data is compensated using nonlinear kernel functions and high-order tensor operations.

[0056] S4. The compensated spectral data were extracted using multi-level recursion to extract features, and the water quality indicators of complex water bodies were predicted using regression models, including total phosphorus, ammonia nitrogen, and total nitrogen.

[0057] Specifically, the specific principle process of the above implementation plan is as follows: the multi-resolution features of the spectrum are extracted through wavelet transform, abnormal fluctuations are identified and suppressed, and at the same time, the noise reduction intensity is dynamically adjusted using an adaptive threshold denoising model to ensure information integrity; after noise reduction, the signal purity of the spectral data is improved; further, through a nonlinear multidimensional gradient compensation algorithm, the influence of turbidity on different wavelengths and times is dynamically adjusted to solve the problem that traditional linear compensation cannot cope with complex water environments; after turbidity compensation, the turbidity effect in the spectral signal is eliminated, so that indicators such as total phosphorus, ammonia nitrogen, and total nitrogen are no longer misled by water turbidity; further, through a multi-level recursive structure, each level of feature extraction model is progressively processed based on the compensated data to ensure the mining and retention of deeper features. Through nonlinear recursive relationships, the complex coupling between features is captured. The characteristic of multi-level recursive extraction is that information goes deeper layer by layer, and richer water quality features are extracted from different dimensions, which significantly improves the resolution of water index analysis, especially the perception of subtle differences in complex environments; further, the extracted multidimensional features are regressed using kernel functions to ensure that nonlinear correlations are fully expressed, especially in different water environments. The segmented regression model can adapt to the different characteristics of the environment. Compared with traditional linear models, segmented nonlinear regression can adapt to sudden changes and nonlinear changes in complex water environments, thereby greatly improving the prediction accuracy of indicators such as total phosphorus, ammonia nitrogen, and total nitrogen; further, the results of recursive feature extraction, kernel regression and dynamic feedback are combined to form the final prediction value. The final prediction value combines multidimensional nonlinear characteristics and real-time optimization results, and can accurately predict indicators such as total phosphorus, ammonia nitrogen, and total nitrogen in complex water environments, greatly improving the accuracy and anti-interference ability of the prediction.

[0058] Furthermore, the step S3 specifically includes the following sub-steps:

[0059] S301. Based on the nonlinear effect of turbidity on the spectrum, the spectral signal is mapped to a high-dimensional space using a kernel method to capture nonlinear features;

[0060] S302. When the characteristics of different wavelengths and time points affect each other, multi-dimensional coupling compensation is performed on the spectral data through a high-order tensor compensation model;

[0061] S303. The nonlinear compensation value output by the kernel method is combined with the tensor compensation term output by the high-order tensor compensation model, a total compensation calculation is performed on the shading degree, and the compensated spectral data is output.

[0062] Specifically, due to the large amount of suspended particulate matter and changes in the optical properties of water (such as turbidity changes) in complex water bodies, the original form of the spectral signal is severely distorted, affecting the accuracy of the indicators; the above implementation plan adopts a multi-dimensional gradient compensation algorithm, and adjusts the compensation coefficient in real time through a nonlinear compensation model according to the wavelength, time and space multi-dimensional data of the spectrum, taking into account the nonlinear effect of turbidity on the spectrum in different bands and times, so as to dynamically adapt to the drastic changes in turbidity in complex water environments. Turbidity compensation reduces the interference of turbid water on the spectral signal, making the spectral recovery closer to the real signal, and through the nonlinear compensation mechanism, effectively cope with the uncertainty of turbidity in complex water bodies, making indicators such as total phosphorus, ammonia nitrogen, and total nitrogen more stable and accurate.

[0063] Furthermore, in step S301, the spectral signal is mapped to a high-dimensional space by a kernel method to capture nonlinear features. The specific process is expressed as follows:

[0064] ;

[0065] Among them, the It represents the turbidity compensation item after compensation, covering the compensation of nonlinear changes of spectral data at different wavelengths and time dimensions. represents the wavelength of the spectral data, Represents the time dimension, that is, the time point of collecting spectral data. Indicates that the wavelength of the spectral data during the turbidity compensation process is The total number of and Indicates the corresponding spectral signal at wavelength and The spectral signal index under Indicates that the wavelength of the spectral data during the turbidity compensation process is The total number of represents the kernel function, represents the spectral data after denoising, represents the pth wavelength point, represents the qth wavelength point, represents the derivative operation, Indicates the adjustment coefficient, used to adjust the wavelength The influence of the second-order derivative at on the compensation results.

[0066] Furthermore, in step S302, performing multi-dimensional coupling compensation on the spectral data by using a high-order tensor compensation model specifically includes the following sub-steps:

[0067] S3021. Model the nonlinear coupling relationship between different wavelengths and time dimensions using a third-order tensor. Express the nonlinear coupling between different wavelengths in complex water bodies using tensor operations, namely:

[0068] ;

[0069] Among them, the represents the turbidity compensation value obtained after calculation by the third-order tensor model. 、 and Respectively represent the wavelength index, the Represents the tensor coefficient of the third-order tensor, reflecting the wavelength 、 and The coupling weight at time t, represents the total number of wavelength dimensions, represents the wavelength of the spectral data, Represents the time dimension, 、 and Represent the i-th wavelength point, j-th wavelength point and k-th wavelength point respectively;

[0070] S3022. Capture the nonlinear variations of spectral signals at different wavelengths and time dimensions through tensor compensation. Further adjust the compensation results based on the adjustment parameters and the influence of the second-order derivatives at different wavelengths and time dimensions, namely:

[0071] ;

[0072] Among them, the Represents the turbidity value after high-order tensor compensation, Represents a scalar parameter used to adjust the weight of the tensor compensation result. represents the derivative operation, Represents the elements of the coefficient matrix, which reflects the coupling strength between the i-th and j-th wavelengths.

[0073] Furthermore, the specific process of step S303 is as follows:

[0074] ;

[0075] Among them, the represents the spectral data after compensation, represents the wavelength of the spectral data, Represents the time dimension, represents the spectral data after denoising, Indicates the turbidity compensation item after compensation. Indicates the turbidity value after high-order tensor compensation.

[0076] Furthermore, the step S4 specifically includes the following sub-steps:

[0077] S401. Perform recursive feature extraction on the pre-processed compensated spectral data, using step-by-step recursion to capture multidimensional nonlinear features under complex water conditions. The step-by-step recursion of multidimensional nonlinear features includes primary feature extraction and secondary feature extraction.

[0078] S402. Predicting indicators based on the extracted features using a nonlinear regression model based on a multidimensional kernel function;

[0079] S403. Integrate the predicted values ​​at all levels calculated by the piecewise nonlinear regression model to generate accurate predictions of water body indicators.

[0080] Specifically, in a complex water environment, the multidimensional features of spectral data often have complex interactions, including coupling relationships between time, space, and environmental factors. It is difficult to reflect these deep information by directly extracting a single feature. The above embodiment mines the multidimensional features in the spectrum layer by layer, recursively processes the feature output of each stage, so that the features are continuously refined and extracted at different scales. Each level of the model not only takes into account the multidimensional information of the current feature, but also retains the information coupling of the previous level, capturing the deep-level correlations and fluctuations in the water environment, and effectively capturing the features in complex water bodies that fluctuate greatly and are closely related to time and environment. It is suitable for the nonlinear and multidimensional changes of feature information in complex water environments, making the analysis of water indicators more comprehensive and accurate, especially when dealing with the interaction of multidimensional features of complex water bodies.

[0081] In addition, data in complex water bodies fluctuate greatly, and water quality indicators change dynamically with the environment. To ensure the continued accuracy of the prediction, the model output needs to be corrected in real time. As a preferred implementation plan of the above embodiment, a dynamic multidimensional feedback mechanism can be introduced. Based on the real-time prediction error, the kernel regression coefficient and the parameters of the feature extraction model can be optimized through the gradient descent method to ensure that the model's adaptive ability in complex water environments is continuously improved. Through the feedback loop, each error will be used to correct the next prediction, so that the model is continuously optimized and gradually approaches the true value.

[0082] Furthermore, in step S401, the first-level feature extraction is specifically as follows: based on the first-level features, recursive feature extraction is performed in combination with features related to the total phosphorus index in the spectral data, namely:

[0083] ;

[0084] Among them, the Represents the first-level feature extraction result, which is used to extract spectral features related to the total phosphorus index. represents the wavelength of the spectral data, Represents the time dimension, Indicates the number of features extracted from the first-level feature. Represents the index of the feature, It represents the weight coefficient of the kth feature in the first-level feature extraction, which is used to characterize the importance of this feature in the prediction of total phosphorus index. Represents the basis function of the kth feature in the first-level feature extraction, which is used to describe the specific method of feature extraction. represents the spectral data after compensation, Represents the adjustment parameters related to the total phosphorus index in the spectral data.

[0085] Furthermore, in step S401, the secondary feature extraction is specifically: recursive feature extraction is performed in combination with the features of the ammonia nitrogen or total nitrogen index in the spectral data, that is:

[0086] ;

[0087] Among them, the Represents the secondary feature extraction result, which is used to extract spectral features related to ammonia nitrogen or total nitrogen indicators. represents the wavelength of the spectral data, Represents the time dimension, Indicates that Indicates the number of features extracted from the secondary feature. Represents the weight coefficient of the kth feature in the secondary feature extraction, Represents the basis function of the kth feature in the secondary feature extraction, Represents the first-level feature extraction result, Indicates the adjustment parameters related to ammonia nitrogen or total nitrogen indicators.

[0088] Furthermore, in step S402, the nonlinear regression model based on the multidimensional kernel function is specifically expressed as:

[0089] ;

[0090] Among them, the Represents the predicted value of the i-th water body index, represents the wavelength of the spectral data, Represents the time dimension, Represents the number of features related to the i-th water body index, Indicates the total number of features related to water body indicators, Represents the nonlinear kernel function of the i-th indicator, Represents the feature extraction result of the i-th water body index, Represents adjustment parameters related to multidimensional environmental parameters, Represents the weight coefficient of the j-th feature.

[0091] Furthermore, the multi-dimensional environmental parameters include: turbidity of water, concentration of particulate matter, size and distribution of suspended matter, and light scattering characteristics.

[0092] Furthermore, in step S403, the specific process of integrating the predicted values ​​is as follows:

[0093] ;

[0094] Among them, the Represents the comprehensive water index prediction value, Indicates the total number of water body indicators, Represents the index of water body index, Represents the predicted value of the i-th water body index, represents the wavelength of the spectral data, Represents the time dimension.

[0095] Furthermore, as a preferred specific implementation of the above embodiment, a water index analysis system applied to complex water bodies is proposed, comprising:

[0096] The data extraction and preprocessing module is used to extract spectral data from complex water bodies. The raw spectral data is preprocessed through adaptive filtering, which adjusts the filter coefficient according to changes in the water environment and outputs preprocessed smoothed spectral data. The preprocessed data is decomposed into different frequency components of the spectral data through multi-scale adaptive wavelet transform and denoised using adaptive thresholding.

[0097] The turbidity compensation module is used to compensate for the turbidity of the spectral data after noise reduction through nonlinear kernel functions and high-order tensor operations;

[0098] An index prediction module is used to extract features from the compensated spectral data through multi-level recursion and predict water body indicators of complex water bodies through regression models, wherein the water body indicators include total phosphorus, ammonia nitrogen and total nitrogen;

[0099] The indicator prediction module specifically includes:

[0100] A nonlinear feature capture unit is used to map the spectral signal to a high-dimensional space by a kernel method based on the nonlinear effect of turbidity on the spectrum, so as to capture the nonlinear features;

[0101] Multidimensional coupling compensation unit, used to perform multidimensional coupling compensation on spectral data through a high-order tensor compensation model when features at different wavelengths and time points affect each other;

[0102] The total compensation calculation unit is used to combine the nonlinear compensation value output by the kernel method with the tensor compensation term output by the high-order tensor compensation model, perform total compensation calculation on the shading, and output the compensated spectral data.

[0103] The indicator prediction module specifically includes:

[0104] A recursive feature extraction unit is used to perform recursive feature extraction on the pre-processed compensated spectral data, and to capture multidimensional nonlinear features under complex water conditions using step-by-step recursion. The step-by-step recursion of multidimensional nonlinear features includes primary feature extraction and secondary feature extraction.

[0105] An indicator regression prediction unit is used to predict indicators based on the extracted features through a nonlinear regression model based on a multidimensional kernel function;

[0106] The prediction integration unit is used to integrate the prediction values ​​of each level calculated by the piecewise nonlinear regression model to generate accurate predictions of water body indicators.

[0107] The foregoing description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein and should not be construed as excluding other embodiments. Rather, the present invention can be used in various other combinations, modifications, and environments and can be modified within the scope of the concept described herein through the above teachings or techniques or knowledge in the relevant field. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention are intended to be protected by the appended claims.

Claims

1. A water index analysis method applied to complex water bodies, characterized in that: The following steps are involved: S1. Extract the spectral data of complex water bodies, pre-process the original spectral data through adaptive filtering, adjust the filter coefficient according to the changes in the water environment, and output the pre-processed smoothed spectral data; S2. The preprocessed data is decomposed into different frequency components of the spectral data by multi-scale adaptive wavelet transform, and denoised by adaptive threshold; S3. The spectral data after noise reduction is compensated for turbidity through nonlinear kernel function and high-order tensor operation; S4. Extracting features from the compensated spectral data through multi-level recursion, and predicting water indicators of complex water bodies through regression models, wherein the water indicators include total phosphorus, ammonia nitrogen and total nitrogen; Wherein, the step S3 specifically includes the following sub-steps: S301. According to the nonlinear effect of turbidity on the spectrum, the spectral signal is mapped to a high-dimensional space by a kernel method to capture nonlinear features; S302. When the characteristics of different wavelengths and time points affect each other, multi-dimensional coupling compensation is performed on the spectral data through a high-order tensor compensation model; S303. Combining the nonlinear compensation value output by the kernel method with the tensor compensation term output by the high-order tensor compensation model, performing a total compensation calculation on the turbidity, and outputting the compensated spectral data; The step S4 specifically includes the following sub-steps: S401. Recursive feature extraction is performed on the pre-processed compensated spectral data, and multi-dimensional nonlinear features under complex water conditions are captured by step-by-step recursion. The step-by-step recursion of multi-dimensional non-linear features includes primary feature extraction and secondary feature extraction; S402. Predicting the index of the extracted features by a nonlinear regression model based on a multidimensional kernel function; S403. Integrate the prediction values ​​at all levels calculated by the piecewise nonlinear regression model to generate accurate predictions of water body indicators.

2. A water body index analysis method applied to complex water bodies as claimed in claim 1, characterized in that: In step S301, the spectral signal is mapped to a high-dimensional space by a kernel method to capture nonlinear features. The specific process is as follows: ； Among them, the It represents the compensation item of turbidity effect after compensation, covering the compensation of nonlinear changes of spectral data in different wavelength and time dimensions. represents the wavelength of the spectral data, represents the time dimension, that is, the time point of collecting spectral data. Indicates that the wavelength of the spectral data during the turbidity compensation process is The total number of and Indicates the corresponding spectral signal at wavelength and The spectral signal index under Indicates that the wavelength of the spectral data during the turbidity compensation process is The total number of represents the kernel function, represents the spectral data after denoising, represents the pth wavelength point, represents the qth wavelength point, represents the derivative operation, Indicates the adjustment coefficient, used to adjust the wavelength The influence of the second-order derivative at on the compensation results.

3. A water body index analysis method applied to complex water bodies as claimed in claim 1, characterized in that: In step S302, performing multi-dimensional coupling compensation on spectral data by using a high-order tensor compensation model specifically includes the following sub-steps: S3021. Model the nonlinear coupling relationship between different wavelengths and time dimensions through third-order tensor, and express the nonlinear coupling between different wavelengths in complex water bodies through tensor operations, namely: ； Among them, the represents the turbidity compensation value obtained after calculation by the third-order tensor model, , and Respectively represent the wavelength index, Represents the tensor coefficient of the third-order tensor, reflecting the wavelength , and The coupling weight at time t is represents the total number of wavelength dimensions, represents the wavelength of the spectral data, Represents the time dimension, , and Respectively represent the i-th wavelength point, the j-th wavelength point and the k-th wavelength point; S3022. Through tensor compensation, the nonlinear changes of spectral signals at different wavelengths and time dimensions are captured, and the compensation results are further adjusted according to the adjustment parameters and the influence of the second-order derivatives at different wavelengths and time dimensions, namely: ； Among them, the Represents the turbidity value after high-order tensor compensation. Represents a scalar parameter used to adjust the weight of the tensor compensation result. represents the derivative operation, Represents the elements of the coefficient matrix, which reflects the coupling strength between the i-th and j-th wavelengths.

4. The water body index analysis method applied to complex water bodies as claimed in claim 1, characterized in that: The specific process of step S303 is as follows: ； Among them, the represents the spectral data after compensation, represents the wavelength of the spectral data, Represents the time dimension, represents the spectral data after denoising, It represents the compensation item of turbidity influence after compensation. Represents the turbidity value after high-order tensor compensation.

5. The water body index analysis method applied to complex water bodies as claimed in claim 1, characterized in that: In step S401, the primary feature extraction is specifically: based on the primary feature, recursive feature extraction is performed in combination with the features related to the total phosphorus index in the spectral data, that is: ； Among them, the Represents the first-level feature extraction result, which is used to extract the spectral features related to the total phosphorus index. represents the wavelength of the spectral data, Represents the time dimension, Indicates the number of features extracted from the first-level feature. Represents the index of the feature, It represents the weight coefficient of the kth feature in the primary feature extraction, which is used to characterize the importance of this feature in the prediction of total phosphorus index. Represents the basis function of the kth feature in the first-level feature extraction, which is used to describe the specific method of feature extraction. represents the spectral data after compensation, Represents the adjustment parameters related to the total phosphorus index in the spectral data.

6. The water body index analysis method applied to complex water bodies as claimed in claim 1, characterized in that: In step S401, the secondary feature extraction is specifically: recursive feature extraction is performed in combination with the features of the ammonia nitrogen or total nitrogen index in the spectral data, that is: ； Among them, the Represents the secondary feature extraction result, which is used to extract spectral features related to ammonia nitrogen or total nitrogen indicators. represents the wavelength of the spectral data, Represents the time dimension, Indicates that Indicates the number of features extracted from the secondary feature. Represents the weight coefficient of the kth feature in the secondary feature extraction. represents the basis function of the kth feature in the secondary feature extraction, Represents the first-level feature extraction result, Indicates the adjustment parameters related to ammonia nitrogen or total nitrogen indicators.

7. The water body index analysis method applied to complex water bodies as claimed in claim 1, characterized in that: In step S402, the nonlinear regression model based on the multidimensional kernel function is specifically expressed as: ； Among them, the represents the predicted value of the i-th water body index, represents the wavelength of the spectral data, Represents the time dimension, represents the number of features related to the i-th water body indicator, Represents the total number of features related to water indicators, Represents the nonlinear kernel function of the i-th index, represents the feature extraction result of the i-th water body index, represents the adjustment parameters related to the multidimensional environmental parameters, Represents the weight coefficient of the jth feature.

8. A water body index analysis method applied to complex water bodies as claimed in claim 7, characterized in that: The multi-dimensional environmental parameters include: turbidity of water, concentration of particulate matter, size and distribution of suspended matter, and scattering characteristics of light.

9. The water body index analysis method applied to complex water bodies as claimed in claim 1, characterized in that: In step S403, the specific process of predictive value integration is expressed as follows: ； Among them, the Represents the comprehensive water index prediction value, Represents the total number of water body indicators, Represents the index of the water body index. represents the predicted value of the i-th water body index, represents the wavelength of the spectral data, Represents the time dimension.

Images