Quantitative analysis method and system for three water color elements coupled with ultraviolet-visible absorption and reflection spectrums

Abstract

The invention relates to an ultraviolet-visible absorption and reflection spectrum coupled water color three-element quantitative analysis method and system, and the method comprises the steps: obtaining a plurality of sample water body reflection signals, and calculating the reflectivity based on the reflection signals; acquiring the total absorption spectrum of the ultraviolet-visible wave band of each sample water body and the absorption spectrum and the concentration of each component of three water color elements; constructing a total absorption spectrum inversion model of the water body based on the reflectivity and the total absorption spectrum of each sample water body; constructing a water color three-element absorption spectrum separation and concentration inversion model based on the absorption spectrum characteristics of each component in the sample water body; and obtaining a reflected signal of a water body to be detected, obtaining a total absorption spectrum of the water body to be detected based on the total absorption spectrum inversion model of the water body, and realizing water color three-element absorption spectrum separation and concentration inversion of the water body to be detected by utilizing the water color three-element absorption spectrum and the concentration inversion model. According to the invention, the problem that CDOM and NAP visible wave band absorption spectrums are difficult to separate is overcome, and high-precision remote sensing observation of water color three-element concentration can be realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of water body monitoring and analysis, and specifically relates to a method and system for quantitatively analyzing three elements of water color by coupling ultraviolet-visible absorption and reflection spectra. Background Art

[0002] As an important bio-optical indicator, water color's dynamic changes carry crucial information about the aquatic environment and are of central importance to ecological research, resource management, and environmental security. The optical characteristics of water are determined by stable water molecules and variable optically active components (OACs). Phytoplankton, colored dissolved organic matter (CDOM), and non-algae particulate matter (NAP) constitute the three key elements of water color. The optical properties and concentration variations of these three components directly influence the spectral characteristics of water, making them key components of remote water color monitoring.

[0003] The current mainstream water color remote sensing algorithms can be divided into two categories: empirical models and semi-analytical models. The empirical model directly establishes the remote sensing reflectance R rs The statistical relationship between the concentration of OACs and the concentration of saturated fatty acids is simple to operate but has significant regional dependence. The semi-analytical model obtains better universality by numerically solving the radiation transfer equation. Among them, the quasi-analytical algorithm (QAA) is a typical representative and can be used to calculate the concentration of saturated fatty acids by R rs Invert the total absorption coefficient a(λ) and separate it into the phytoplankton absorption coefficient a ph (λ), the sum of the absorption coefficients of NAP and CDOM a dg However, since this algorithm assumes that NAP and CDOM have similar exponential decay spectral shapes, it cannot achieve accurate separation of the two.

[0004] Technical bottlenecks primarily arise from two aspects: First, existing satellite sensors generally lack the ultraviolet band (the characteristic absorption region of CDOM), making it difficult to effectively distinguish the spectral contributions of CDOM and NAP using only visible light. Second, while laboratory / in situ UV-visible absorption spectroscopy can accurately capture water absorption characteristics, its temporal and spatial coverage is orders of magnitude lower than that of remote sensing observations. This technological disconnect between "high-precision point measurements" and "large-scale surface observations" severely restricts the accuracy and efficiency of dynamic water environment monitoring. Summary of the Invention

[0005] To address the existing difficulty in separating CDOM and NAP in remote water color sensing, this paper proposes a method and system for quantitatively analyzing the three water color elements by coupling UV-visible absorption spectroscopy with remotely sensed reflectance. This approach aims to leverage deep neural networks to learn to predict UV-visible absorption spectra from readily available reflectance data, effectively integrating the spatiotemporal observation capabilities of remote sensing with the accurate spectral information provided by laboratory or in situ spectroscopic measurements. Furthermore, the absorption spectra are decomposed using a model with clear physical meaning, aiming to effectively separate the three water color elements, particularly CDOM and NAP. This invention aims to provide a new approach for the precise inversion of OACs components in water.

[0006] The present invention adopts the following technical solutions.

[0007] In one aspect, the present invention discloses a method for quantitatively analyzing the three elements of water color by coupling ultraviolet-visible absorption and reflectance spectra, characterized in that the method comprises the following steps:

[0008] Step 1: Use an optical sensor to obtain reflection signals from various sample water bodies and calculate the reflectivity R based on the reflection signals. rs ;

[0009] Step 2: Use a UV-visible absorption spectrometer to obtain the total absorption spectrum a(λ) of the UV-visible band of each sample water body and the absorption spectrum and concentration of each component of the three elements of water color;

[0010] Step 3: Based on the reflectance R of each sample water body rs (λ) and the total absorption spectrum a(λ), and construct the total absorption spectrum inversion model of water body;

[0011] Step 4: Based on the absorption spectrum characteristics of each component in the sample water, a water color three-element absorption spectrum separation and concentration inversion model is constructed;

[0012] Step 5: Obtain the reflection signal of the water body to be tested, obtain the total absorption spectrum a(λ) of the water body to be tested based on the total absorption spectrum inversion model of the water body, and use the absorption spectrum and concentration inversion model of the three elements of water color to achieve the absorption spectrum separation and concentration inversion of the three elements of water color of the water body to be tested.

[0013] More preferably,

[0014] In step 1, the reflectivity R rs Calculated using the following formula:

[0015]

[0016] Among them L t is the measured water surface reflection signal of the sample water, L skyIt is the signal of sky light reflected from the water surface, L p is the reflection signal of the standard white plate, and r is the reflectivity between the water and air interfaces.

[0017] More preferably,

[0018] In step 3, a total absorption spectrum inversion model of water body is constructed based on machine learning;

[0019] Calculate R rs The logarithm of R rs The first derivative and R rs The first derivative of the logarithm of R rs 、R r The logarithm of s and the first-order derivatives of the two are used as input samples to train the total absorption spectrum inversion model of water bodies.

[0020] More preferably,

[0021] The constructed absorption spectrum prediction model adopts a feedforward neural network architecture with multiple hidden layers. The network hyperparameters are automatically determined through Bayesian optimization combined with K-fold cross-validation. The Adam optimizer is used in the training process to minimize the root mean square error (RMSE) between the predicted spectrum and the target spectrum.

[0022] More preferably,

[0023] Construct a water color three-element absorption spectrum separation and concentration inversion model, specifically including:

[0024] Subtract the pure water absorption a from the total absorption spectrum a(λ) w (λ) and chlorophyll absorption a ph (λ), and the sum of the absorption coefficients of CDOM and NAP components a is obtained dg (λ), and the UV-visible absorption spectra of CDOM and NAP were used to construct a dg (λ) separation model to achieve absorption spectrum separation and concentration inversion of CDOM and NAP.

[0025] More preferably,

[0026] The total absorption spectrum coefficient a(λ) of water is expressed as:

[0027] a(λ)=a w (λ)+a ph (λ)+a g (λ)+a d (λ)

[0028] Among them, a w (λ) is the absorption coefficient of pure water, a ph (λ) is the absorption coefficient of chlorophyll components, a g(λ) is the absorption coefficient of CDOM components, a d (λ) is the absorption coefficient of NAP component.

[0029] More preferably,

[0030] In step 4, a ph (λ) is calculated as follows:

[0031] a ph =C chl ·a ph *(λ)

[0032] C chl =f(R rs (λ1),R rs (λ2),…,R rs (λn)

[0033] Among them, C chl is the chlorophyll component concentration, a ph * (λ) is the absorption spectrum of chlorophyll per unit concentration, R rs (λ1), R rs (λ2), R rs (λn) are the remote sensing reflectances at wavelengths of λ1, λ2 and λn respectively, f is the chlorophyll concentration inversion model based on remote sensing reflectance, and the selection range of f is OC2, OC3, OC4, and OCI models.

[0034] More preferably,

[0035] When f is the OC3 model,

[0036]

[0037] λ1 is the wavelength covering the secondary absorption peak of chlorophyll a, λ2 is the wavelength of the transition zone between the chlorophyll absorption valley and the suspended matter scattering enhancement, and λ3 is the wavelength band reflecting the backscattering of suspended matter; R rs (λ1), R rs (λ2), R rs (λ3) are the remote sensing reflectances at wavelengths of λ1, λ2, and λ3, respectively. a and b are the first and second parameter values ​​of the model, respectively, which are calculated by coupling the measured chlorophyll concentration with the remote sensing reflectance.

[0038] More preferably,

[0039] In step 4, a dg The (λ) separation model is constructed by the following formula:

[0040] a dg (λ)=a g (λ)+a d(λ)

[0041] a g (λ)=C CDOM ·a g * (λ)

[0042]

[0043] a d (λ)=C NAP ·a d * (λ)

[0044]

[0045] Among them, a g (λ) is the absorption coefficient of CDOM components, a d (λ) is the absorption coefficient of NAP component, C CDOM Indicates the concentration of CDOM in the water sample to be tested, λ is the wavelength, C NAP is the concentration of NAP in the water sample to be tested, a g * (λ) represents the concentration-normalized absorption spectrum of CDOM components, a d * (λ) is the concentration-normalized absorption spectrum of NAP components, λ0 is the reference wavelength, and a d * (λ0) is the absorption coefficient at the reference wavelength, S is the spectral slope of the NAP absorption spectrum, A i is the amplitude of the i-th Gaussian function, W i is the width of the i-th Gaussian function, E is the photon energy, E i is the central energy position of the i-th Gaussian function.

[0046] More preferably,

[0047] Each parameter in the absorption spectrum separation and concentration inversion model of the three elements of water color is estimated using the nonlinear least squares method.

[0048] In another aspect, the present invention discloses a method and system for quantitatively analyzing the three elements of water color by coupling ultraviolet-visible absorption and reflectance spectra. The analysis system includes one or more processors and one or more programs, and is characterized by:

[0049] One or more programs are stored in the one or more memories and are configured to be executed by the one or more processors, and the one or more programs include instructions for executing the aforementioned water color three-element quantitative analysis method.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] This paper proposes and verifies a new method that can effectively separate the absorption contributions of CDOM and NAP with highly overlapping spectral features in water bodies, and successfully constructs and preliminarily verifies a water body component calculation framework that couples ultraviolet-visible absorption and reflectance spectra by combining deep neural networks and physical models.

[0052] The present invention uses deep neural network to obtain the remote sensing reflectivity R rs The full-band absorption spectrum a(λ) containing key ultraviolet information was predicted, which provided the necessary spectral basis for the subsequent differentiation of CDOM and NAP, overcoming the limitation of traditional methods that lack ultraviolet information.

[0053] The present invention applies the predicted absorption spectrum to a physically meaningful decomposition model, achieving high-precision inversion of chlorophyll a, CDOM, and NAP concentrations (chlorophyll aR 2 =0.95,CDOMR 2 =0.97,NAPR 2 =0.99). This result highlights the significant advantages of this coupling method in solving the long-standing bottleneck problem of the three elements of water color (especially the separation of CDOM and NAP).

[0054] The physical model used in this paper describes the two differently: using a combination of Gaussian functions to precisely characterize CDOM, while using exponential decay combined with scattering correction to describe the optical properties of NAP. This differentiated parameterization method, combined with the full spectrum information including the ultraviolet band provided by the neural network, enables the model to more effectively "deconstruct" a dg (λ), thus achieving a g (λ) and a d (λ) Further separation of contributions. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 Schematic diagram of the process of the quantitative analysis method of the three elements of water color by coupling ultraviolet-visible absorption and reflectance spectra of the present invention.

[0056] Figure 2 This is a diagram of the experimental setup for measuring reflectance spectrum in Example 1.

[0057] Figure 3 The total absorption coefficient spectra of some simulated water samples under different NAP, chlorophyll a, and CDOM concentration gradients. (a) Chlorophyll a is 0.2 μg·L -1 , NAP is 1 mg·L -1 (b) Chlorophyll a is 1 μg·L -1 , NAP is 5 mg·L -1(c) Chlorophyll a is 10 μg·L -1 , NAP is 20 mg·L -1 .

[0058] Figure 4 The remote sensing reflectance spectra of some simulated water samples under different NAP, chlorophyll a, and CDOM concentration gradients. (a) Chlorophyll a is 0.2 μg·L -1 , NAP is 1 mg·L -1 (b) Chlorophyll a is 1 μg·L -1 , NAP is 5 mg·L -1 (c) Chlorophyll a is 10 μg·L -1 , NAP is 20 mg·L -1 .

[0059] Figure 5 Evaluation of the neural network prediction results. (a) Comparison of the mean of the actual absorption coefficient and the mean of the predicted absorption coefficient; (b) Scatter plot of the actual and predicted absorption coefficients based on six typical wavelengths in the training set; (c) Scatter plot of the actual and predicted absorption coefficients based on six typical wavelengths in the test set.

[0060] Figure 6 The performance of the coupled method in retrieving the concentrations of the three water color elements on an independent test set is verified. (a) Chlorophyll a concentration; (b) CDOM concentration; (c) NAP (kaolin) concentration. DETAILED DESCRIPTION

[0061] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. The embodiments described in this application are only part of the embodiments of the present invention, not all of them. Based on the spirit of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0062] like Figure 1 As shown, the present invention discloses a method for quantitatively analyzing the three elements of water color by coupling ultraviolet-visible online absorption and reflectance spectroscopy, the method comprising the following steps:

[0063] Step 1: Use an optical sensor to obtain reflection signals from various sample water bodies and calculate the reflectivity R based on the reflection signals. rs ;

[0064] Obtain the reflection signal of the water sample to be tested, and calculate the reflectivity R based on the reflection signal rs ;

[0065]

[0066] Among them L t It is the measured water surface reflection signal of the water sample to be tested, L sky It is the signal of sky light reflected from the water surface, L p is the reflection signal of the standard white plate, r is the reflectivity between the water and air interfaces;

[0067] Step 2: Use a UV-visible absorption spectrometer to obtain the total absorption spectrum a(λ) of the UV-visible band of each sample water body and the absorption spectrum and concentration of each component of the three elements of water color;

[0068] Step 3: Based on the reflectance R of each sample water body rs (λ) and the total absorption spectrum a(λ), and construct the total absorption spectrum inversion model of water body;

[0069] In order to realize the mapping from reflection spectrum to absorption spectrum, the intrinsic optical properties of water absorption spectrum are obtained through the apparent optical signal, and a machine learning model simulating the reflection spectrum and absorption spectrum data of water samples is constructed as the total absorption spectrum inversion model of water body. rs Spectrum, R rs The logarithm of the spectrum and the first-order derivatives of the two are used as input features, and the input features are z-score standardized to realize the calculation from reflectance spectrum to absorption spectrum.

[0070] The machine learning inversion model includes but is not limited to random forest, support vector machine, neural network, etc.

[0071] The present invention preferably builds an absorption spectrum prediction model based on a feedforward neural network model. The model body adopts a feedforward neural network architecture containing multiple hidden layers, integrates batch normalization, ReLU activation and Dropout and other technologies, and is intended to effectively learn the complex nonlinear mapping relationship between the reflection spectrum and the absorption spectrum and improve generalization ability. The hyperparameters of the network are automatically determined by Bayesian optimization in combination with K-fold cross validation to ensure that the model performance is optimal. The training process uses Adam optimizer, with the root mean square error (RMSE) between the predicted spectrum and the target spectrum as the goal, and combines the early stopping strategy to prevent overfitting. The standardized predicted spectrum of the output is subjected to denormalization and optional Gaussian smoothing to obtain the final predicted absorption spectrum, and will be used for the absorption spectrum decomposition and component inversion described in the follow-up.

[0072] Step 4: Based on the absorption spectral characteristics of each component in the sample water, a water color three-element absorption spectral separation and concentration inversion model is constructed; specifically including:

[0073] Subtract the pure water absorption a from the total absorption spectrum a(λ) w (λ) and chlorophyll absorption a ph (λ), and the sum of the absorption coefficients of CDOM and NAP components a is obtaineddg (λ), and the UV-visible absorption spectra of CDOM and NAP were used to construct a dg (λ) separation model to achieve absorption spectrum separation and concentration inversion of CDOM and NAP.

[0074] Since the main optically active components of water include pure water, chlorophyll (i.e., phytoplankton component), CDOM, and NAP, the absorption spectrum of the water sample can be represented by the additive absorption spectrum model of the four:

[0075] a(λ)=a w (λ)+a ph (λ)+a g (λ)+a d (λ)

[0076] Among them, aw(λ) is the absorption of pure water, aph(λ) is the absorption of chlorophyll component, ag(λ) is the absorption of CDOM component, and ad(λ) is the absorption of NAP component.

[0077] The absorption spectrum coefficient a(λ) of the optically active substance in the water sample to be tested is obtained by deducting the baseline absorption of pure water from the total absorption spectrum coefficient a(λ) of the water sample to be tested. nw (λ), expressed as:

[0078] a nw (λ)=a ph (λ)+a g (λ)+a d (λ)

[0079] Among them, a ph (λ) is the absorption coefficient of chlorophyll components, a g (λ) is the absorption coefficient of CDOM components, a d (λ) is the absorption coefficient of NAP component;

[0080] The absorption coefficient of chlorophyll components is calculated based on the absorption spectrum characteristics of chlorophyll components, and the absorption spectrum coefficient a of optically active substances is obtained. nw The sum of the absorption coefficients of CDOM and NAP components is obtained by separating and deducting the absorption contribution of chlorophyll components from (λ) dg (λ):

[0081] a dg (λ)=a g (λ)+a d (λ)

[0082] a ph (λ) is calculated as follows:

[0083] a ph (λ)=C chl ·a ph*(λ)

[0084] C chl =f(R rs (λ1),R rs (λ2),…,R rs (λn)

[0085] Among them, C chl is the concentration of chlorophyll components in the water sample to be tested, a ph * (λ) is the absorption spectrum of chlorophyll component per unit concentration, R rs (λ1), R rs (λ2), R rs (λn) are the remote sensing reflectances at wavelengths λ1, λ2, and λn, respectively. f is the chlorophyll concentration inversion model based on remote sensing reflectance. The selection range of f is OC2, OC3, OC4, OCI model, etc. Take the OC3 model as an example:

[0086]

[0087] Among them, R rs (λ1), R rs (λ2), R rs (λ3) is the remote sensing reflectance at wavelengths of λ1, λ2, and λ3, respectively. a and b are the first and second parameter values ​​of the model, respectively, and are calculated by coupling the measured chlorophyll concentration with the remote sensing reflectance. The values ​​of λ1, λ2, and λ3 are usually selected from bands such as around 660 nm (the band where the absorption and fluorescence peaks of chlorophyll a are located), 690-730 nm (the band that is greatly affected by CDOM and NAP absorption), and 740-760 nm (the band where water body reflection is enhanced and affected by suspended matter backscattering). The specific values ​​are selected according to actual conditions.

[0088] a dg The (λ) separation model is constructed by the following formula:

[0089] a dg (λ)=a g (λ)+a d (λ)

[0090] a g (λ)=C CDOM ·a g * (λ)

[0091]

[0092] a d (λ)=C NAP ·a d * (λ)

[0093]

[0094] Among them, a g (λ) is the absorption coefficient of CDOM components, a d (λ) is the absorption coefficient of NAP component, C CDOM Indicates the concentration of CDOM in the water sample to be tested, λ is the wavelength, C NAP is the concentration of NAP in the water sample to be tested, a g * (λ) represents the concentration-normalized absorption spectrum of CDOM components, a d * (λ) is the concentration-normalized absorption spectrum of NAP components, λ0 is the reference wavelength, and a d * (λ0) is the absorption coefficient at the reference wavelength, S is the spectral slope of the NAP absorption spectrum, A i is the amplitude of the i-th Gaussian function, W i is the width of the i-th Gaussian function, E is the photon energy, E i is the central energy position of the i-th Gaussian function.

[0095] In a preferred embodiment of the present invention, a g * The absorption spectrum of (λ) in the range of 200-800 nm is accurately described by three Gaussian functions:

[0096]

[0097] Where E is the photon energy in eV; λ is the wavelength in nm; E i is the central energy position of the i-th Gaussian function, A i is the amplitude of the i-th Gaussian function, W i is the width of the i-th Gaussian function.

[0098] Each relevant parameter in the fitting formula is calculated using the continuous iterative nonlinear least squares method to obtain the UV-visible absorption spectrum information of the water body, thereby achieving the separation of CDOM and NAP.

[0099] Step 5: Obtain the reflection signal of the water body to be tested, obtain the total absorption spectrum a(λ) of the water body to be tested based on the total absorption spectrum inversion model of the water body, and use the absorption spectrum and concentration inversion model of the three elements of water color to achieve the absorption spectrum separation and concentration inversion of the three elements of water color of the water body to be tested.

[0100] The technical solution of the present invention is simulated in the laboratory and the technical effect is verified.

[0101] 1. Simulation experiment on the quantitative analysis method of the three elements of water color by coupling UV-visible absorption spectroscopy with remote sensing reflectance

[0102] 1.1 Experimental simulated water sample configuration

[0103] In order to construct and verify the technical solution of the present invention, laboratory simulation methods were used to prepare simulated water samples with different OACs concentration gradients that can represent a wide range of water types from clear water to turbid water, oligotrophic to eutrophic. Commercial chlorophyll a standards, internationally accepted Suwannee River Natural Organic Matter (SRNOM) standards and analytical grade kaolin powder were selected as simulants of phytoplankton, colored dissolved organic matter (CDOM) and non-algae particulate matter (NAP), respectively. Ultrapure water was used as the base, and by accurately adding different amounts of chlorophyll a, SRNOM and kaolin, simulated water samples were prepared, the concentrations were calibrated, and their reflectance and UV-visible absorption spectra were measured simultaneously. Kaolin and chlorophyll a were graded using concentration gradients, with the gradient range being 0.1-10 μg·L -1 and 1-20 mg·L -1 , SRNOM uses its absorption coefficient at 443nm (a g (443)) was graded with a gradient range of 0.05-1.6 m -1 .

[0104] 1.2 Reflection and absorption spectrum measurement and processing

[0105] Build a reflectance spectrum measurement device in the laboratory ( Figure 2 ). A Newport 94061A simulated sunlight source was used, a 1.4L black barrel was used as a container to hold the simulated water sample, and a QE65-pro spectrometer from Ocean Optics was used to measure the reflectance spectrum of the water sample. Finally, a simulated water sample L with a spectral range of 350-800nm ​​and a spectral resolution of 1nm was obtained. t , L skt , L p , and use formula (1) to calculate the reflectivity R of the water sample rs .

[0106]

[0107] Among them, L t The spectrum instrument probe receives the simulated water sample surface reflection signal, L sky It is the signal of sky light reflected from the water surface, L p is the reflection signal of the standard white plate, and r is the reflectivity between the water and air interface.

[0108] The absorption spectrum of water samples was measured using a Lambda 850 UV-visible spectrophotometer produced by Perkin Elmer with a measurement range of 250–800 nm and a spectral resolution of 1 nm.

[0109] 1.3 Reflection spectrum to absorption spectrum mapping method

[0110] In order to realize the mapping from reflection spectrum to absorption spectrum, the intrinsic optical properties of water absorption spectrum are obtained through the apparent optical signal, and a neural network model simulating the reflection spectrum and absorption spectrum data of water samples is constructed using the deep learning method. rs Spectrum, R rs The logarithm of the spectrum and the first-order derivatives of the two are used as input features. The input features are z-score standardized to construct a machine learning prediction model for the absorption spectrum, realizing the calculation from the reflection spectrum to the absorption spectrum.

[0111] Among them, the machine learning inversion model includes but is not limited to random forest, support vector machine, neural network, etc.

[0112] The present invention preferably builds an absorption spectrum prediction model based on a feedforward neural network model. The model body adopts a feedforward neural network architecture containing multiple hidden layers, integrates batch normalization, ReLU activation and Dropout and other technologies, and is intended to effectively learn the complex nonlinear mapping relationship between the reflection spectrum and the absorption spectrum and improve generalization ability. The hyperparameters of the network are automatically determined by Bayesian optimization in combination with K-fold cross validation to ensure that the model performance is optimal. The training process uses Adam optimizer, with the root mean square error (RMSE) between the predicted spectrum and the target spectrum as the goal, and combines the early stopping strategy to prevent overfitting. The standardized predicted spectrum of the output is subjected to denormalization and optional Gaussian smoothing to obtain the final predicted absorption spectrum, and will be used for the absorption spectrum decomposition and component inversion described in the follow-up.

[0113] 1.4 Absorption spectrum analysis method

[0114] The main optically active components of water include pure water, chlorophyll, CDOM and NAP. Therefore, the absorption spectrum of the water sample can be represented by the additive absorption spectrum model of the four:

[0115] a(λ)=a w (λ)+a ph (λ)+a g (λ)+a d (λ) (2)

[0116] Among them, a w (λ) is the absorption of pure water, a ph (λ) is the absorption of chlorophyll components, a g (λ) is the absorption of CDOM components, ad (λ) is the absorption of NAP component.

[0117] Since the experiment uses a double-beam spectrophotometer to measure the absorption with ultrapure water as the background reference, the baseline absorption of pure water has been deducted before measuring the solution absorption. Therefore, the absorption spectrum of the water sample obtained in the experiment can be expressed as formula (3):

[0118] a nw (λ)=a ph (λ)+a g (λ)+a d (λ) (3)

[0119] Chlorophyll a has significant absorption peaks in the blue-violet and red light regions, while the absorption spectra of CDOM and NAP have similar characteristics, both of which decay exponentially with increasing wavelength. Based on the significant differences in the absorption spectra of chlorophyll a, CDOM and NAP, the absorption contribution of chlorophyll a can be effectively separated and deducted from the total absorption. ph (λ) can be expressed as equations (4)-(5):

[0120] a ph =C chl ·a ph * (λ) (4)

[0121]

[0122] Among them, C chl is the chlorophyll concentration; a ph * (λ) is the absorption spectrum of chlorophyll per unit concentration; R rs (λ1), R rs (λ2), R rs (λ3) is the remote sensing reflectance at wavelengths of λ1, λ2, and λ3, respectively. Here, the values ​​of λ1, λ2, and λ3 are 660, 710, and 748 nm, respectively. a and b are model parameter values, which can be calculated by coupling the measured chlorophyll concentration with the remote sensing reflectance. Here, the values ​​of a and b are 289.2 and 31.57, respectively.

[0123] Absorption of CDOM componentsa g (λ) is expressed as Equation 6-7:

[0124] a g (λ)=C CDOM ·a g * (λ)·ε (6)

[0125]

[0126] Among them, CCDOM represents the concentration of CDOM in the laboratory solution; ε represents the interference correction term of NAP; γ is the maximum degree of attenuation of the CDOM absorption spectrum, which describes the maximum inhibition ratio of NAP on CDOM absorption; δ is the attenuation rate of the CDOM absorption spectrum, which describes the change rate of the inhibition of the CDOM absorption spectrum when the NAP concentration increases; C NAP is the concentration of NAP (kaolin) in the laboratory solution. a g * (λ) represents the concentration-normalized CDOM absorption spectrum. a g * The absorption spectrum of (λ) in the range of 200-800 nm can be accurately described by three Gaussian functions:

[0127]

[0128] Where E is the photon energy in eV; λ is the wavelength in nm; E i is the central energy position of the i-th Gaussian function, A i is the amplitude of the i-th Gaussian function, W i is the width of the i-th Gaussian function.

[0129] Absorption of NAP componentsa d The (λ) coefficient is expressed as formula (10):

[0130] a d (λ)=C NAP ·a d * (λ) (10)

[0131] Among them, a d * (λ) is the concentration-normalized absorption spectrum, which is described by an exponential function. However, since the laboratory did not use quantitative filter technology when measuring the UV-visible absorption spectrum of simulated water samples, a d * The (λ) spectrum also includes the scattered part produced by NAP, which is a d * (λ) is a correction term, so in the present invention, a d * (λ) can be expressed as:

[0132]

[0133] Where λ0 is the reference wavelength, usually 440nm; a d * (λ0) is the absorption coefficient at the reference wavelength; S is the spectral slope of the absorption spectrum, which determines the decay rate of absorption with wavelength.* (λ) represents a d * The scattering correction term of (λ) can be approximately described by equation (12):

[0134]

[0135] Where m is the scattering correlation coefficient, which describes the scattering intensity; n is the wavelength index of scattering, which describes the relationship between the scattering intensity and wavelength; β is the saturation factor, which is used to simulate the saturation phenomenon of scattering under high concentration of NAP.

[0136] According to the UV-visible absorption spectra and concentration data of laboratory simulated water samples, equations (3)-(12) are combined and the model parameters are solved using the nonlinear least squares method.

[0137] 1.5 Method accuracy verification

[0138] To evaluate the proposed method for quantitatively analyzing the three elements of water color by coupling UV-visible online absorption and reflectance spectroscopy, the accuracy of the method was verified using laboratory-measured reflectance, absorption spectra, and concentrations. First, the total absorption spectrum was calculated based on the laboratory-measured reflectance spectrum, and the accuracy of the total absorption spectrum inversion was verified. Then, based on the inverted total absorption spectrum and the constructed absorption spectrum analysis model, the inverted total absorption spectrum was decomposed into chlorophyll, CDOM, and NAP absorption spectra. The sum of the decomposed component spectra was compared with the measured total absorption spectrum to evaluate the spectral analysis accuracy of the absorption spectrum analysis model for the three elements of water color. Finally, the constructed absorption spectrum analysis model was used to solve the concentrations of the three components, and the accuracy of the inverted three-component concentrations was evaluated.

[0139] 2 Analysis of experimental results

[0140] 2.1 Laboratory reflection and absorption spectrum measurement results

[0141] Figure 3 The total absorption coefficients of some simulated water samples under different NAP, chlorophyll a, and CDOM concentration gradients are shown. Figure 3 From left to right, the three parts are: (a) Chlorophyll a is 0.2 μg·L -1 , NAP is 1 mg·L -1 (b) Chlorophyll a is 1 μg·L -1 , NAP is 5 mg·L -1 (c) Chlorophyll a is 10 μg·L -1 , NAP is 20 mg·L -1 .

[0142] Depend on Figure 3It can be seen that the total absorption coefficient at different NAP, chlorophyll a, and CDOM concentration gradients exhibits a nearly exponential decay with increasing wavelength. While NAP and chlorophyll a concentrations remain constant, the total absorption coefficient increases systematically across the entire spectrum with increasing CDOM concentration, with a particularly pronounced increase in the shortwave band. While chlorophyll a and CDOM concentrations remain constant, the total absorption coefficient rises across the entire wavelength range with increasing NAP concentration. While NAP and CDOM concentrations remain constant, the total absorption coefficient also changes with increasing chlorophyll a concentration, with characteristic peaks in the blue and red regions also increasing. However, these changes in absorption morphology are masked by the strong absorption of CDOM and NAP.

[0143] Figure 4 The remote sensing reflectance of some simulated water samples under different NAP, chlorophyll a, and CDOM concentration gradients is shown. Figure 4 From left to right, the three parts are: (a) Chlorophyll a is 0.2 μg·L -1 , NAP is 1 mg·L -1 (b) Chlorophyll a is 1 μg·L -1 , NAP is 5 mg·L -1 (c) Chlorophyll a is 10 μg·L -1 , NAP is 20 mg·L -1 .

[0144] Figure 4 It shows that NAP concentration mainly affects the overall intensity of reflectivity. As NAP concentration increases, its backscattering also increases, R rs CDOM concentration primarily affects the magnitude and spectral slope of remote sensing reflectance in the blue-green band. At fixed background concentrations of NAP and chlorophyll a, as CDOM concentration increases, its strong absorption in the blue-violet region leads to a significant decrease in Rrs in this region and a steeper spectral slope. As chlorophyll a concentration increases, absorption dips appear in the blue (approximately 440 nm) and red (approximately 675 nm) regions. Simultaneously, a reflectance peak appears in the near-infrared band (approximately 700 nm), which increases with increasing chlorophyll a concentration.

[0145] 2.2 Remote sensing reflectance analysis of the three components of water color

[0146] Figure 5 The inversion results of the total absorption spectrum using the deep neural network estimation absorption spectrum model based on the reflectance spectrum are shown. Figure 5From left to right, the three parts are: (a) comparison of the mean of the actual absorption coefficient and the mean of the predicted absorption coefficient; (b) scatter distribution of the actual absorption coefficient and the predicted absorption coefficient based on six typical wavelengths in the training set; (c) scatter distribution of the actual absorption coefficient and the predicted absorption coefficient based on six typical wavelengths in the test set.

[0147] Figure 5 Part (a) shows the comparison between the actual absorption coefficient mean and the predicted absorption coefficient mean in the wavelength range of 250-800nm. Figure 5 Parts (b) and (c) show scatter plots of the predicted and actual values ​​of six typical wavelengths in the training and test sets.

[0148] Depend on Figure 5 As can be seen from part (a), the average absorption spectrum predicted by the model is highly consistent with the actual measured average spectrum over the entire wavelength range of 250-800nm. The standard deviation of the predicted results is also close to the standard deviation of the actual values, which shows that the waveform also has a good ability to predict the degree of dispersion of the data around the mean. Figure 5 From parts (b) and (c), we can see that the predicted coefficients of the six typical wavelengths show good linear correlation with the actual absorption coefficients; the scatter distribution of the training set and the test set is consistent, indicating that the model has no obvious overfitting and has a certain generalization ability.

[0149] Table 1 further shows that at typical wavelengths, the R 2 The RMSE values ​​are generally low, and the RMSE of the test set is close to that of the training set, indicating stable model performance and low prediction error. Overall, the model successfully learns and predicts absorption coefficient spectra, with reliable results and good potential for practical application.

[0150] Table 1 Evaluation of neural network prediction results of six typical bands

[0151]

[0152] The concentrations of the three water color elements were inverted based on the constructed absorption spectrum analytical model, and the accuracy was verified with the measured concentrations. The results are as follows: Figure 6 shown. Figure 6 From left to right, (a) chlorophyll a concentration; (b) CDOM concentration; (c) NAP (kaolin) concentration are shown. The results show that this method can achieve high-precision inversion of chlorophyll a, CDOM, and NAP concentrations, and the determination coefficient R 2The results were 0.95, 0.97, and 0.99, respectively, and the root mean square error (RMSE) was 0.72 μg·L -1 , 0.25mg·L -1 , 0.72mg·L -1 .

[0153] The present invention also seeks to protect a water color three-element quantitative analysis system coupled with ultraviolet-visible online absorption and reflectance spectroscopy, the analysis system comprising one or more processors and one or more programs, wherein the one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, the one or more programs comprising instructions for executing the water color three-element quantitative analysis method.

[0154] The present disclosure may be a system, method and / or computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0155] A computer-readable storage medium can be a tangible device that can hold and store instructions for use by an instruction execution device. A computer-readable storage medium can be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination thereof. As used herein, a computer-readable storage medium is not to be construed as a transient signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through an electrical wire.

[0156] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions to be stored in the computer-readable storage medium in each computing / processing device.

[0157] The computer program instructions for performing the operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, and conventional procedural programming languages ​​such as "C" language or similar programming languages. Computer-readable program instructions may be executed entirely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., utilizing an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may be personalized by utilizing the state information of the computer-readable program instructions. The electronic circuit may execute the computer-readable program instructions, thereby realizing various aspects of the present disclosure.

[0158] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.

Claims

1. A method for quantitative analysis of the three elements of water color by coupling ultraviolet-visible absorption and reflectance spectra, characterized in that: The method comprises the following steps: Step 1: Use an optical sensor to obtain reflection signals from various sample water bodies and calculate the reflectivity R based on the reflection signals. rs ; Step 2: Use a UV-visible absorption spectrometer to obtain the total absorption spectrum a(λ) of the UV-visible band of each sample water body and the absorption spectrum of each component of the three water color elements, and simultaneously measure the concentration of the three water color elements; Step 3: Based on the reflectance R of each sample water body rs (λ) and the total absorption spectrum a(λ), and construct the total absorption spectrum inversion model of water body; Step 4: Based on the absorption spectrum characteristics of each component in the sample water, a water color three-element absorption spectrum separation and concentration inversion model is constructed; Step 5: Obtain the reflection signal of the water body to be tested, obtain the total absorption spectrum a(λ) of the water body to be tested based on the total absorption spectrum inversion model of the water body, and use the absorption spectrum and concentration inversion model of the three elements of water color to achieve the absorption spectrum separation and concentration inversion of the three elements of water color of the water body to be tested.

2. The method for quantitative analysis of the three elements of water color according to claim 1, characterized in that: In step 1, the reflectivity R rs Calculated using the following formula: Among them L t is the measured water surface reflection signal of the sample water, L sky It is the signal of sky light reflected from the water surface, L p is the reflection signal of the standard white plate, and r is the reflectivity between the water and air interfaces.

3. The method for quantitative analysis of the three elements of water color according to claim 1, characterized in that: In step 3, a total absorption spectrum inversion model of water body is constructed based on machine learning; Calculate R rs The logarithm of R rs The first derivative and R rs The first derivative of the logarithm of R rs 、R rs The logarithm of and the first-order derivatives of the two are used as input samples to train the total absorption spectrum inversion model of water bodies.

4. The method for quantitative analysis of the three elements of water color according to claim 3, characterized in that: The constructed absorption spectrum prediction model adopts a feedforward neural network architecture with multiple hidden layers. The network hyperparameters are automatically determined through Bayesian optimization combined with K-fold cross-validation. The Adam optimizer is used in the training process to minimize the root mean square error (RMSE) between the predicted spectrum and the target spectrum.

5. The method for quantitative analysis of the three elements of water color according to claim 1, characterized in that: Construct a water color three-element absorption spectrum separation and concentration inversion model, specifically including: Subtract the pure water absorption a from the total absorption spectrum a(λ) w (λ) and chlorophyll absorption a ph (λ), and the sum of the absorption coefficients of CDOM and NAP components a is obtained dg (λ), and the UV-visible absorption spectra of CDOM and NAP were used to construct a dg (λ) separation model to achieve absorption spectrum separation and concentration inversion of CDOM and NAP.

6. The method for quantitative analysis of the three elements of water color according to claim 5, characterized in that: The total absorption spectrum coefficient a(λ) of water is expressed as: a(λ)=a w (λ)+a ph (λ)+a g (λ)+a d (l) Among them, a w (λ) is the absorption coefficient of pure water, a ph (λ) is the absorption coefficient of chlorophyll components, a g (λ) is the absorption coefficient of CDOM components, a d (λ) is the absorption coefficient of NAP component.

7. The method for quantitative analysis of the three elements of water color according to claim 5, characterized in that: In step 4, a ph (λ) is calculated as follows: a ph =C chl ·a ph * (l) C chl =f(R rs (λ1),R rs (λ2),…,R rs (λn)) Among them, C chl is the chlorophyll component concentration, a ph * (λ) is the absorption spectrum of chlorophyll per unit concentration, R rs (λ1), R rs (λ2), R rs (λn) are the remote sensing reflectances at wavelengths of λ1, λ2 and λn respectively, f is the chlorophyll concentration inversion model based on remote sensing reflectance, and the selection range of f is OC2, OC3, OC4, and OCI models.

8. The method for quantitative analysis of the three elements of water color according to claim 7, characterized in that: When f is the OC3 model, λ1 is the wavelength covering the secondary absorption peak of chlorophyll a, λ2 is the wavelength of the transition zone between the chlorophyll absorption valley and the suspended matter scattering enhancement, and λ3 is the wavelength band reflecting the suspended matter backscattering; R rs (λ1), R rs (λ2), R rs (λ3) are the remote sensing reflectances at wavelengths of λ1, λ2, and λ3, respectively. a and b are the first and second parameter values ​​of the model, respectively, which are calculated by coupling the measured chlorophyll concentration with the remote sensing reflectance.

9. The method for quantitative analysis of the three elements of water color according to claim 5, characterized in that: In step 4, a dg The (λ) separation model is constructed by the following formula: a dg (λ)=a g (λ)+a d (l) a g (λ)=C CDOM ·a g * (l) a d (λ)=C NAP ·a d * (l) Among them, a g (λ) is the absorption coefficient of CDOM components, a d (λ) is the absorption coefficient of NAP component, C CDOM Indicates the concentration of CDOM in the water sample to be tested, λ is the wavelength, C NAP is the concentration of NAP in the water sample to be tested, a g * (λ) represents the concentration-normalized absorption spectrum of CDOM components, a d * (λ) is the concentration-normalized absorption spectrum of NAP components, λ0 is the reference wavelength, and a d * (λ0) is the absorption coefficient at the reference wavelength, S is the spectral slope of the NAP absorption spectrum, A i is the amplitude of the i-th Gaussian function, W i is the width of the i-th Gaussian function, E is the photon energy, E i is the central energy position of the i-th Gaussian function.

10. The method for quantitative analysis of the three elements of water color according to claim 5 or 9, characterized in that: Each parameter in the absorption spectrum separation and concentration inversion model of the three elements of water color is estimated using the nonlinear least squares method.

11. A method and system for quantitatively analyzing the three elements of water color by coupling ultraviolet-visible absorption and reflectance spectroscopy, the analysis system comprising one or more processors and one or more programs, characterized in that: One or more programs are stored in the one or more memories and configured to be executed by the one or more processors, and the one or more programs include instructions for executing the water color three-element quantitative analysis method according to any one of claims 1 to 10.

Images