A method for remotely detecting distribution of CDOM fluorescence and non-fluorescent substances in water body

Abstract

The application discloses a water body CDOM fluorescence and non-fluorescent substance distribution remote sensing detection method, survey points are arranged in a target river basin, CDOM is detected by using an unmanned aerial vehicle hyperspectral detector, and FDOM is detected by using a fluorescence laser radar; the concentrations of CDOM and FDOM of some survey points are measured; a CDOM concentration inversion model is established by combining the CDOM detected by the hyperspectral detector with the measured CDOM concentration; the CDOM data of all survey points are obtained by combining the inversion model with the hyperspectral data of the survey points; a FDOM concentration inversion model is established by combining the fluorescence data detected by the LIF technology with the measured FDOM concentration data; the FDOM data of all survey points are obtained by combining the inversion model with the fluorescence spectral data of the survey points; the NFDOM concentration is obtained by combining the CDOM concentration data and the FDOM concentration data; and the concentration distribution of FDOM and NFDDOM is simulated by analyzing the data.

Description

TECHNICAL FIELD

[0001] The present application relates to a water body CDOM fluorescence and non-fluorescent substance distribution remote sensing detection method, belonging to the fields of remote sensing and marine science. BACKGROUND

[0002] Chromophoric Dissolved Organic Matter (CDOM) is a part of Dissolved Organic Matter (DOM) that has a very obvious absorption spectrum in the short wave band from blue to ultraviolet, which can significantly affect the underwater light field and the biogeochemical cycle of carbon, nitrogen, phosphorus and other biogenic elements in the water body. CDOM is mainly composed of soluble organic mixtures such as humus, fulvic acid and aromatic polymer components, and is the largest dissolved organic carbon reservoir in the water body, playing an important role in the global carbon cycle. CDOM has important significance for studying water body carbon cycle and responding to global greenhouse effect.

[0003] Part of the organic matter in CDOM has certain fluorescence characteristics, that is, when short-wavelength light is used as excitation light, the fluorescent substance in CDOM will absorb the excitation light and emit fluorescence longer than the excitation light wavelength. CDOM can be divided into Fluorescent Dissolved Organic Matter (FDOM) and Non-Fluorescent Dissolved Organic Matter (NFDOM) according to whether it has fluorescence characteristics. At present, the research on CDOM is focused on the inversion model of CDOM concentration, and the use of three-dimensional fluorescence spectrum to explore the composition of FDOM in various water bodies to analyze the source characteristics of CDOM. The source of CDOM is affected by surrounding land use, climate change, industrial and agricultural pollution emissions, etc., and is highly related to the water quality of the water body, which is an important indicator of water body condition. Exploring the composition of FDOM through three-dimensional fluorescence spectrum technology can analyze the source of CDOM, but after FDOM absorbs ultraviolet light, most of it is converted into NFDOM through fluorescence quenching, and only a small part has fluorescence characteristics. Therefore, exploring the distribution of FDOM and NFDOM in the water body can determine the best position for monitoring the water quality of the target water area through three-dimensional fluorescence spectrum technology, and also can analyze the contribution of FDOM and NFDOM to water pollution respectively, which is helpful for monitoring and evaluating the water quality, ecological environment, etc. of the water body. In view of the current situation that there is no method for detecting the distribution of FDOM and NFDOM in the water body, the present application provides a water body CDOM fluorescence and non-fluorescent substance distribution remote sensing detection method combining hyperspectral remote sensing and LIF technology. SUMMARY

[0004] The present application is to solve the above technical problems, provide a kind of water CDOM fluorescence and non-fluorescent substance distribution remote sensing detection method, can explore the distribution of FDOM and NFDOM in water, can determine the best position of target water area water quality monitored by three-dimensional fluorescence spectroscopy technology, also can be analyzed respectively FDOM and NFDOM contribution to water pollution, help to monitor and assess water quality, ecological environment etc.

[0005] To solve the above technical problems, the technical solution adopted by the present application is:

[0006] A kind of water CDOM fluorescence and non-fluorescent substance distribution remote sensing detection method, according to the area of investigation region, target area is grid processing, the entire region is divided into multiple areas 1m 2 Grid, as blank data point, and the current region boundary is limited, the data of entire target water area region blank grid is handled, after data processing, the point of same concentration is connected to represent the continuous distribution of FDOM and NFDOM and gradually changing quantity characteristics;Meanwhile, through geographic software, the investigation point is arranged according to the actual situation of target basin, the FDOM concentration and NFDOM concentration actually monitored by investigation point are divided into several levels from small to large, and the contour distribution diagram of FDOM and NFDOM concentration is drawn, to simulate the distribution of FDOM and NFDOM in entire target basin;

[0007] The specific method for data processing of entire target water area region blank grid data is:

[0008] If the estimation of variable in random field is expressed as linear system containing random error ε, then the best linear unbiased prediction BLUP of Gauss-Markov theorem is expressed as selecting linear system parameter, so that the variance of estimation value And observation value Y reaches minimum, that is

[0009]

[0010]

[0011] In the formula: x0 It is unknown point;{x1,x2,…,x n}It is the sample of random field;λ It is weight coefficient;

[0012] In static stationary random field under isotropic assumption, mathematical expectation μ (x) It is irrelevant to position, and covariance is the function of distance|h| Between points;Usually, variance function C is unknown, and needs to use variation function as approximation, at this time, variation function γ It is only related to the distance between points, that is

[0013] E[Y(x)]=μ,

[0014] var[Y(x)-Y(x+h)]=2|C(0)-C(h)|=2γ(|h|),

[0015] define {Y(x1),Y(x2),…,Y(x n )} as n survey point FDOM or CDOM concentration values, the expression and variance are as follows:

[0016]

[0017]

[0018] In the formula: Y is the estimation of FDOM and NFDOM concentration values Y at unknown point x0;n is the number of spatial correlation longitude and latitude point pairs at x0, C(x i ,x j ) is the covariance function between points x i and x j , according to the BLUP theory, the unbiased estimation condition is that the sum of all weight coefficients satisfies:

[0019]

[0020] Therefore, the following solving function can be constructed using the Lagrange multiplier method:

[0021]

[0022]

[0023] Solve the equation set to obtain the weight coefficient, and substitute the obtained weight into the formula to obtain the unbiased estimation of FDOM and NFDOM concentration

[0024] The further improvement of the technical scheme of the application is that the specific method for monitoring the FDOM concentration and the NFDOM concentration of the survey point is:

[0025] Step one: use the unmanned aerial vehicle hyperspectral detector to detect CDOM and the fluorescent laser radar to detect FDOM, simultaneously perform hyperspectral image data and laser radar point cloud data matching to obtain the spatial correspondence of the two kinds of data;And simultaneously carry out water quality monitoring work of sampling points, select part of the survey points to carry out actual sampling, and measure the CDOM and FDOM concentration thereof;

[0026] Step two: combine the hyperspectral data detected by the hyperspectral detector with the measured CDOM concentration data to establish a CDOM concentration inversion model;obtain all the survey point CDOM data by combining the hyperspectral data of the survey points through the inversion model;

[0027] Step three: combine the fluorescence data detected by the LIF technology with the measured FDOM concentration data to establish an FDOM concentration inversion model; obtain the FDOM data of all survey points by combining the fluorescence spectrum data of the survey points through the inversion model;

[0028] Step four: combine the CDOM concentration data and the FDOM concentration data of the survey points to obtain the NFDOM concentration of all survey points.

[0029] The further improvement of the technical scheme of the present application is that in step one, the hyperspectral detector and the fluorescence lidar system are connected as a whole framework, and the framework is mounted below the unmanned aerial vehicle.

[0030] The ground control console sends instructions to the microcontroller of the unmanned aerial vehicle flight control through wireless communication, and the microcontroller controls the hyperspectral detector, the fluorescence lidar system and the IMU / GPS system to synchronously collect data; the unmanned aerial vehicle control system synchronously transmits the POS data of the fluorescence lidar system to the hyperspectral detector, so that the hyperspectral detector and the fluorescence lidar system have the same position information and time information when detecting, and the collected data is transmitted back to the ground station computer through wireless connection; wherein, the POS data is the position data and attitude data of the unmanned aerial vehicle.

[0031] The further improvement of the technical scheme of the present application is that in step one, the resampling method is used to match the spatial resolution of the spectral data obtained by remote sensing detection and the point cloud data generated by the lidar system, so as to realize the spatial consistency of the hyperspectral data and the fluorescence lidar point cloud data.

[0032] If the image resolution of the spectral image data is lower than the spatial resolution of the lidar point cloud data, the spectral image data is resampled by an interpolation method; on the contrary, if the image resolution of the spectral image data is higher than the spatial resolution of the lidar point cloud data, the lidar point cloud data is resampled by an up-sampling method.

[0033] After resampling, the lidar point cloud image is converted into a 2D image to obtain a point cloud grid, the intensity value of each pixel in the point cloud grid corresponds to the average intensity value of each foot point in the point cloud grid, and the point cloud grid is used as a reference image for spectral image data registration; the spectral image data matched with the waveband of the lidar detector in the spectral image is used as a template image for image data matching, so that the spectral image data and the fluorescence lidar point cloud data can be spatially corresponding.

[0034] Further improvement of the technical scheme of the present application is that in the step one, the CDOM is detected by the hyperspectral detector through the hyperspectral sensor carried by the unmanned aerial vehicle to shoot the hyperspectral image of the investigation point, and the digital quantization value DN value of each wave band of the target position can be directly obtained from the hyperspectral image, the DN value is a value quantified by the hyperspectral sensor receiving radiation, which is related to the quantization depth, has no unit and no actual meaning, and is converted into reflectance data through radiation calibration, and specifically, the radiation calibration method is that a calibration plate is placed under the lens of the hyperspectral detector, the ambient light intensity can be obtained through the calibration plate, so that the water surface reflectance is calculated.

[0035] Further improvement of the technical scheme of the present application is that in the step one, the process that the fluorescence lidar system detects the water body FDOM is that the laser of the fluorescence lidar system emits laser of a specific wavelength, the laser is reflected by the emission mirror of the system and finally hits the water surface at a specific angle, and the FDOM in the water body emits fluorescence longer than the excitation light wavelength after absorbing the excitation light; the fluorescence signal emitted by the FDOM is received by the telescope of the fluorescence lidar system, passes through the collimating lens, passes through the optical filter to the photosensitive sensor and is received by the system.

[0036] Further improvement of the technical scheme of the present application is that in the step one, the actual sampling depth is 0.5m below the water surface, the brown bottle constant temperature box is used for low-temperature and light-proof preservation, and the CDOM concentration and the FDOM concentration of the sample are measured by the ultraviolet-luminometer and the spectrophotometer respectively.

[0037] Further improvement of the technical scheme of the present application is that in the step two, the specific operation of establishing the CDOM inversion model is that the original spectral information of the target position with the wavelength of 400-1000nm is extracted from the obtained hyperspectral image, and the original spectral data is converted into reflectance spectral data through radiation calibration.

[0038] The radiation calibration method is that the hyperspectral reflectance information can be obtained through the reflected light intensity and the incident light intensity, the reflected light intensity can be considered as the DN value of the current environment, the incident light intensity can be considered as the incident light intensity of the calibration plate, because the reflectance of the calibration plate and the reflected light intensity are known, so the incident light intensity of the current environment can be obtained; the hyperspectral reflectance calculation formula is as follows:

[0039]

[0040] In the formula, ρ1 is the reflectance of the calibration plate, DN1 is the DN value of the calibration plate, DN is the DN value of the investigation point, and ρ is the reflectance of the investigation point. t t

[0041] ​​In spectral analysis, the intensity values of spectral data will be different due to the influence of different sample concentrations, temperatures, pressures and other factors. In order to eliminate the interference of these factors, the spectral data is normalized to more accurately analyze the relationship between the measured concentration data and the spectral remote sensing reflectance. The normalization formula is as follows:

[0042]

[0043] In the formula: λ i is the ith wavelength; the value range of i is 400-1000; L(λ i ) is the initial value of the reflectance of wavelength λ i ; n is the number of wavelengths contained in the wavelength range; L N (λ i ) is the normalized reflectance of wavelength λ i .

[0044] The spectral internal segmentation method is adopted, the spectrum is divided into a wave band every 10 nm, the average reflectance in the wave band is taken as the reflectance data of the whole wave band, and through the wave band traversal, the iterative algorithm is adopted to perform Pearson analysis on the remote sensing reflectance data of each wave band and the CDOM concentration, construct a correlation coefficient graph, select the wave band reflectance ratio factor with high correlation for modeling, the higher the absolute value of the Pearson correlation coefficient, the more the wavelength can represent the CDOM concentration change, and the expression of the correlation coefficient R is as follows:

[0045]

[0046] In the formula: x i is the reflectance of a wave band of the ith survey point; is the average reflectance of all survey points in the same wave band; y i is the concentration of the water quality parameter of the ith survey point; is the average concentration of all survey points; and R is the correlation coefficient value.

[0047] Taking the selected highest correlation wave band reflectance as the independent variable and the CDOM concentration value as the dependent variable, a linear inversion formula for inverting the CDOM concentration is established:

[0048] C CDOM = k1×R rs + α,

[0049] In the formula: R rs is the remote sensing reflectance data of the highest correlation wave band; k1 is the coefficient of the inversion formula; and α is the intercept of the inversion formula, and the values of k1 and α are determined by combining the measured data analysis;

[0050] By combining the CDOM inversion model with hyperspectral data from the survey sites, the CDOM concentration numbers for all survey sites can be obtained through inversion.

[0051] A further improvement of the technical solution of the present invention is as follows: In step three, the specific operation of establishing the FDOM inversion model is as follows: When short wavelength light is used as the excitation wavelength, FDOM will absorb and release longer wavelength light; Fn(355) is defined as the maximum fluorescence intensity of the emission wavelength Em in the range of 440-470nm when the excitation wavelength Ex is 355nm, which represents the relative concentration level of humic substances in FDOM; humic substances are the components with relatively high content in FDOM, which are used to characterize the concentration of FDOM. The water body Fn(355) parameter is usually used to characterize the water body FDOM concentration.

[0052] Based on the linear relationship between Fn(355) and FDOM concentration, the inversion algorithm based on the empirical model of fluorescence peak intensity yields the following formula:

[0053] C FDOM =k2×Fn(355)+β,

[0054] In the formula: Fn(355) is the maximum fluorescence intensity of the emission wavelength Em in the range of 440-470 nm when the excitation wavelength Ex is 355 nm; k2 is the coefficient of the inversion formula; β is the intercept of the inversion formula. The values ​​of k2, β, and α are determined by combining measured data.

[0055] By combining the FDOM inversion model with the fluorescence spectral data of the survey sites, the FDOM concentrations at all survey sites can be obtained.

[0056] A further improvement to the technical solution of this invention lies in the following: In step four, the formula for calculating the NFDOM concentration is as follows:

[0057] C NFDOM =C CDOM -C FDOM ,

[0058] In the formula: C CDOM For CDOM concentration data at the survey sites; C FDOM For FDOM concentration data at the survey sites; C NFDOM To collect NFDOM concentration data at the survey sites.

[0059] The technological advancements achieved by this invention due to the adoption of the above technical solutions are as follows:

[0060] A remote sensing method for detecting the distribution of CDOM (fluorescent and non-fluorescent substances) in water bodies is proposed. This method can explore the distribution of FDOM and NFDOM in water bodies, determine the optimal location for monitoring the water quality of target water areas using three-dimensional fluorescence spectroscopy, and also separately analyze the contribution of FDOM and NFDOM to water pollution. This method is helpful for monitoring and assessing the water quality and ecological environment of water bodies. Attached Figure Description

[0061] Figure 1 This is a flowchart of a remote sensing method for detecting the distribution of CDOM fluorescence and non-fluorescent substances in water bodies.

[0062] Figure 2 This is a flowchart illustrating the structure and data matching process of an unmanned aerial vehicle (UAV) detection system.

[0063] Figure 3 This is a technology roadmap for LIF technology to remotely sense FDOM.

[0064] Figure 4 This is a technical roadmap for CDOM (Continuous Doppler Oscillator) in water bodies using hyperspectral remote sensing.

[0065] Figure 5 It is a technology roadmap for simulating FDOM and NFDOM distributions. Detailed Implementation

[0066] To make the technical means, creative features, objectives and effects of the present invention easy to understand, the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them.

[0067] Combination Figures 1-5 The specific implementation method adopts the following technical solution: a remote sensing detection method for the distribution of CDOM fluorescence and non-fluorescent substances in water bodies combining hyperspectral and LIF technologies, combined with... Figure 1 As shown, it includes the following steps:

[0068] Step 1: Set up survey points in the target watershed, use UAV hyperspectral detectors to detect CDOM and fluorescent lidar to detect FDOM, and match hyperspectral image data with lidar point cloud data to obtain the spatial correspondence between the two types of data; at the same time, carry out water quality monitoring at the sampling points, select some survey points for actual sampling, and measure their CDOM and FDOM concentrations.

[0069] Step 2: Combine the hyperspectral data detected by the hyperspectral detector with the measured CDOM concentration data to establish a CDOM concentration inversion model;

[0070] Step three: combining the fluorescence data detected by the LIF technology and the measured FDOM concentration data, an FDOM concentration inversion model is established;

[0071] Step four: combining the CDOM concentration data and the FDOM concentration data of the survey points, the NFDOM concentration of all survey points is obtained.

[0072] Step five: analyzing the data, the concentration distribution of FDOM and NFDDOM in the entire target basin is simulated.

[0073] In the step one, the survey points are arranged by combining the geographic software with the actual situation of the target basin. The survey point position should be arranged as much as possible to cover the entire survey area, and the water quality and hydrological characteristics (such as the water inlet area, the water outlet area, the deep water area, the shallow water area, the shore area, etc.) of the water body such as the sea, the lake, the reservoir, etc. should be marked.

[0074] In the embodiment, the hyperspectral detector and the fluorescence lidar system are connected as a whole framework, and the framework is mounted under the unmanned aerial vehicle. At the same time, in order to realize the spatial consistency of the hyperspectral data and the fluorescence lidar point cloud data, the hyperspectral image data and the fluorescence lidar point cloud data are matched by resampling. The structure of the unmanned aerial vehicle detection system and the data matching process are shown in Figure 2 .

[0075] The ground control console sends instructions to the microcontroller of the flight control of the unmanned aerial vehicle through wireless communication, and the microcontroller controls the hyperspectral detector, the fluorescence lidar system and the IMU / GPS system to synchronously collect data. The unmanned aerial vehicle control system synchronously transmits the POS data (position data and attitude data of the unmanned aerial vehicle) of the fluorescence lidar system to the hyperspectral detector, so that the hyperspectral detector and the fluorescence lidar system have the same position information and time information when detecting. At the same time, the collected data is transmitted back to the ground station computer through wireless connection.

[0076] Further, in order to obtain the spatial correspondence of the hyperspectral image data and the fluorescence lidar point cloud data, the hyperspectral image data and the fluorescence lidar point cloud data are spatially matched. The present application uses the resampling method to match the spatial resolution of the hyperspectral image data and the point cloud data generated by the fluorescence lidar system.

[0077] If the image resolution of the spectral image data is lower than the spatial resolution of the lidar point cloud data, the spectral image data is resampled by interpolation method; on the contrary, if the image resolution of the spectral image data is higher than the spatial resolution of the lidar point cloud data, the lidar point cloud data is resampled by upsampling method.

[0078] The laser radar point cloud image is converted into a 2D image after resampling to obtain a point cloud grid. The intensity value of each pixel in the point cloud grid corresponds to the average intensity value of each foot point in the point cloud grid, and the point cloud grid is taken as a reference image for spectral image data registration. The spectral image data matching the waveband of the laser radar detector in the spectral image is taken as a template image. Image data matching is performed to enable spatial correspondence between the spectral image data and the fluorescent laser radar point cloud data.

[0079] The resampling principle of hyperspectral image data by interpolation is as follows:

[0080] Let (x, y) be the pixel coordinates of the interpolation point in the resampled hyperspectral image, and the neighborhood point coordinates be (i, j). Let the gray value of (i, j) be f(x, y) on the (i, j) coordinate plane.

[0081] The cubic convolution interpolation function is defined as:

[0082]

[0083] In the formula: s is the distance between the interpolation point and the neighborhood point.

[0084] f(x, y) = f(i+x d ,j+y d ),

[0085] In the formula: x d , y d is the difference between the horizontal and vertical coordinates of the interpolation point and the neighborhood point.

[0086] The principle of upsampling of the point cloud data of the fluorescent laser radar system is as follows:

[0087] For the root point a of the point cloud data of the fluorescent laser radar system, set the neighborhood search sphere radius R, and there are N neighborhood points. A neighborhood fitting plane is established by using the moving least squares method.

[0088]

[0089] In the formula: p a is the neighborhood point density of the root point a.

[0090] Set the point cloud density threshold p0. When p a < p0, project the root point a of the point cloud data to the fitting plane of a uniformly and randomly, and the intensity value of the projection point is determined by the intensity value of its neighborhood point. Until the neighborhood point density of the root point a satisfies p a ≥ p0; obtain the resampled point cloud image; repeat the search of the point cloud root point and perform the above operation to complete the resampling of the laser radar point cloud.

[0091] Further, the process of image data registration comprises:

[0092] The SIFT algorithm is used to find the spectral image T closest to the laser radar detector waveband as a template image; R is a reference image, and an affine transformation space registration target function is established

[0093]

[0094] In the formula: b is a corresponding matched feature point pair in T and R, T(b) is the corresponding pixel of b on T, R(b) is the corresponding pixel of b on R, is an affine transformation acting on T, and a similarity measure between R;

[0095] The affine transformation is:

[0096]

[0097] In the formula: is a rotation matrix, is a translation matrix;

[0098] The optimization algorithm is used to optimize the target function to obtain the best rotation matrix and the best translation matrix, and the best rotation matrix and the best translation matrix are used to act on all collected spectral image data, so that the image data registration is completed.

[0099] Further, in order to establish the inversion model, part of the points are actually sampled. The actual sampling points are as evenly distributed as possible in the entire basin. The sampling depth is 0.5m below the water surface, the sample is placed in a brown bottle into a constant temperature box for low temperature and light protection, and is taken back to the laboratory. The CDOM concentration and FDOM concentration of the sample are measured by ultraviolet-luminometer and spectrophotometer respectively.

[0100] Further, in the step two, the CDOM inversion model is established by combining the measured data and the hyperspectral data, and the process of establishing the CDOM inversion model is shown in Figure 3 The original spectral information of the target position wavelength in 400-1000nm is extracted from the obtained hyperspectral image, and the original spectral data is converted into reflectivity spectral data by radiation calibration

[0101] The radiation calibration method is: the hyperspectral reflectivity information can be obtained by the reflected light intensity and the incident light intensity, the reflected light intensity can be considered as the DN value of the current environment, and the incident light intensity can be considered as the incident light intensity of the calibration board. Because the reflectivity and the reflected light intensity of the calibration board are known, the incident light intensity of the current environment can be obtained. The calculation formula of the hyperspectral reflectivity is as follows:

[0102]

[0103] wherein ρ t , DN t are the reflectance and DN value to be converted, ρ1 is the reflectance of the calibration plate, and DN1 is the DN value of the calibration plate;

[0104] Further, in spectral analysis, the intensity values of spectral data will be different due to the influence of different sample concentrations, temperatures, pressures and other factors. In order to eliminate the interference of these factors, the spectral data is normalized to more accurately analyze the relationship between the measured concentration data and the spectral remote sensing reflectance. The normalization formula is as follows:

[0105]

[0106] wherein λ i is the ith wavelength; the value range of i is 400-1000; L(λ i ) is the initial value of the reflectance of the wavelength λ i ; n is the number of wavelengths contained in the wavelength range; L N (λ i ) is the normalized reflectance of the wavelength λ i .

[0107] The spectral internal segmentation method is adopted, the spectrum is segmented into a wave band every 10 nm, and the average reflectance in the wave band is taken as the reflectance data of the entire wave band. Through wave band traversal, an iterative algorithm is adopted to perform Pearson analysis on the remote sensing reflectance data of each wave band and the CDOM concentration, to construct a correlation coefficient graph, and to select the wave band reflectance ratio factor with high correlation for modeling. The higher the absolute value of the Pearson correlation coefficient is, the more the wavelength can represent the CDOM concentration change. The expression of the correlation coefficient R is as follows:

[0108]

[0109] wherein x i is the reflectance of a wave band of the ith survey point; is the average reflectance of all survey points in the same wave band; y i is the concentration of the water quality parameter of the ith survey point; is the average concentration of all survey points; and R is the correlation coefficient value;

[0110] The wave band reflectance with the highest correlation is selected as the independent variable, and the CDOM concentration value is selected as the dependent variable, to establish a linear inversion formula for inverting the CDOM concentration:

[0111] C CDOM = k1×R rs + α,

[0112] wherein R rsThe data represents the remote sensing reflectance of the band with the highest correlation; k1 represents the coefficients of the inversion formula; α represents the intercept of the inversion formula. The values ​​of k1 and α are determined by combining measured data analysis.

[0113] By combining the CDOM inversion model with hyperspectral data from the survey sites, the CDOM concentration numbers for all survey sites can be obtained through inversion.

[0114] Furthermore, in step three, an FDOM inversion model is established by combining fluorescence data and measured data. The principle of LIF technology for detecting FDOM is based on its characteristics: when short-wavelength light is used as the excitation wavelength, FDOM absorbs and emits longer-wavelength light. Fn(355) is defined as the maximum fluorescence intensity of the emission wavelength Em in the 440-470nm range when the excitation wavelength Ex is 355nm, representing the relative concentration level of humic substances in FDOM. Humic substances are relatively abundant components in FDOM and can be used to characterize the concentration of FDOM; the Fn(355) parameter is typically used to characterize the concentration of FDOM in water.

[0115] Existing research has found a linear relationship between Fn(355) and FDOM concentration. Based on the empirical model inversion algorithm for fluorescence peak intensity, the formula is:

[0116] C FDOM =k2×Fn(355)+β,

[0117] In the formula: Fn(355) is the maximum fluorescence intensity of the emission wavelength Em in the range of 440-470 nm when the excitation wavelength Ex is 355 nm; k2 is the coefficient of the inversion formula, and β is the intercept of the inversion formula. The values ​​of k2, β, and α are determined by combining measured data.

[0118] By combining the FDOM inversion model with the fluorescence spectral data of the survey sites, the FDOM concentrations at all survey sites can be obtained.

[0119] Furthermore, in step four, by combining the CDOM and FDOM concentration data from the survey sites, the NFDOM concentration for all survey sites is obtained. The formula for calculating the NFDOM concentration is as follows:

[0120] C NFDOM =C CDOM -C FDOM ,

[0121] In the formula: C CDOM For CDOM concentration data at the survey sites; C FDOM For FDOM concentration data at the survey sites; C NFDOM To collect NFDOM concentration data at the survey sites.

[0122] Finally, in step five, the distribution of FDOM and NFDOM in the whole target basin is simulated based on the actual monitoring data of the survey points. According to the area of the survey region, the target region is subjected to grid processing, the whole region is divided into a plurality of grids with an area of 1 m 2 , as blank data points, and the current region boundary is taken as a limit. The blank grid data of the whole target water region is subjected to interpolation processing. Figure 5 The flow chart for drawing the distribution of FDOM and NFDOM in the target basin.

[0123] The specific method is as follows:

[0124] If the estimation of the variable in the random field is expressed as a linear system containing a random error ε, then the best linear unbiased prediction (BLUP) of the Gauss-Markov theorem can be expressed as selecting the linear system parameters so that the variance of the estimation value and the observed value Y is minimized, that is,

[0125]

[0126]

[0127] In the formula, x0 is an unknown point; {x1, x2, …, x n} is a sample of the random field; and λ is a weight coefficient.

[0128] In a static stationary random field under the isotropic assumption, the mathematical expectation μ(x) is irrelevant to the position, and the covariance is only a function of the distance |h| between points. Generally, the variance function C is unknown, and a variogram needs to be used as an approximation, at this time, the variogram γ is also only related to the distance between points, that is,

[0129] E[Y(x)] = μ,

[0130] var[Y(x)-Y(x+h)] = 2|C(0)-C(h)| = 2γ(|h|).

[0131] Define {Y(x1), Y(x2), …, Y(x n )} as the FDOM or CDOM concentration values of n survey points. The expression and variance are as follows:

[0132]

[0133]

[0134] In the formula, is the estimation of the FDOM and NFDOM concentration values Y at an unknown point x0; n is the number of spatial correlation longitude and latitude point pairs that x0 has, C(x i , x j) is the covariance function between points x i and x j According to the BLUP theory, the unbiased estimation condition is that the sum of all weight coefficients satisfies:

[0135]

[0136] Therefore, the following solving function can be constructed using the Lagrange multiplier method and the equation group is obtained:

[0137]

[0138]

[0139] The weight coefficients are obtained by solving the equation group, and the obtained weight is substituted into the formula to obtain the unbiased estimation of the FDOM and NFDOM concentrations

[0140] After the interpolation is completed, the points with the same concentration can be connected to represent the continuous distribution and gradually changing quantity characteristics of the FDOM and NFDOM. Meanwhile, the FDOM concentration and the NFDOM concentration are divided into several levels from small to large, and the contour distribution map of the FDOM and NFDOM concentrations is drawn.

[0141] Although the present application is described herein with reference to specific implementation examples, it should be understood that these embodiments are only examples of the principles and applications of the present application, and therefore, it should be recognized that many modifications can be made to the exemplary embodiments and other arrangements can be proposed without departing from the spirit and scope of the present application as defined in the appended claims. It should be recognized that the various original rights described herein can be different from those described in the original claims and can be combined with the methods of the dependent claims. It should also be recognized that the features described with respect to the embodiments can be used in other described examples.

Claims

1. A method for detecting the distribution of CDOM fluorescence and non-fluorescent substances in a water body by remote sensing, characterized in that: According to the area of the investigation region, the target region is grid processed, the whole region is divided into multiple grids with an area of 1m 2 , as blank data points, and the current region boundary is taken as a limit, the whole target water region blank grid data is processed, after the data processing is completed, the points with the same concentration are connected to represent the continuous distribution and gradually changing quantity characteristics of FDOM and NFDOM, NFDOM represents non-fluorescent dissolved organic matter; meanwhile, through geographical software, the investigation points are arranged according to the actual situation of the target basin, the FDOM concentration and NFDOM concentration actually monitored by the investigation points are divided into several levels from small to large, the contour distribution diagram of FDOM and NFDOM concentration is drawn, and the distribution of FDOM and NFDOM in the whole target basin is simulated. The specific method for data processing of the blank grid data of the whole target water area is as follows: If the estimate of the variable in the random field is expressed as a linear system containing a random error ε, then the best linear unbiased prediction BLUP of the Gauss-Markov theorem is expressed as choosing the parameters of the linear system so that the variance of the estimate and the observation Y is minimized, i.e. where: x0is an unknown point; {x1, x2, …, x n} are samples of the random field; λ is a weight coefficient; In a static stationary random field under the isotropic assumption, the mathematical expectation μ(x) is irrelevant to the position, and the covariance is a function of the distance |h| between points; usually, the variance function C is unknown, and a variation function needs to be used as an approximation, at this time, the variation function γ is also only related to the distance between points, that is E[Y(x)]=μ, var[Y(x)-Y(x+h)]=2|C(0)-C(h)|=2γ(|h|), Define {Y(x1),Y(x2),…,Y(x)} n Let} represent the FDOM or CDOM concentration values ​​at n survey points, with the expression and variance as follows: where: Y is the estimated concentration value of FDOM and NFDOM at unknown point x0; n is the number of spatially correlated longitude-latitude point pairs at x0, C(x i ,x j ) is the covariance function between points x i and x j , according to the BLUP theory, the unbiased estimation condition is that the sum of all weight coefficients satisfies: Therefore, the following solving function is constructed by using the Lagrange multiplier method: Solving the equations yields the weight coefficients, which are substituted into the equation to obtain the unbiased estimates of the concentrations of FDOM and NFDOM The specific method for monitoring the FDOM concentration and the NFDOM concentration at the survey points is as follows: Step one: the CDOM and the FDOM are detected by using the unmanned aerial vehicle hyperspectral detector and the fluorescence laser radar, the hyperspectral image data and the laser radar point cloud data are matched to obtain the spatial correspondence of the two kinds of data; and the water quality monitoring work of the sampling points is simultaneously carried out, and the CDOM and the FDOM concentrations of part of the survey points are actually sampled; Step two: the CDOM concentration inversion model is established by combining the hyperspectral data detected by the hyperspectral detector and the actually measured CDOM concentration data; the CDOM data of all the survey points are obtained by combining the hyperspectral data of the survey points through the inversion model; Step three: the FDOM concentration inversion model is established by combining the fluorescence data detected by the LIF technology and the actually measured FDOM concentration data; the FDOM data of all the survey points are obtained by combining the fluorescence spectral data of the survey points through the inversion model; Step four: the NFDOM concentration of all the survey points is obtained by combining the CDOM concentration data and the FDOM concentration data of the survey points.

2. The method according to claim 1, wherein the method is characterized by: In the step one, the hyperspectral detector and the fluorescence laser radar system are connected as an integral framework, and the framework is hung below the unmanned aerial vehicle; The ground control console sends instructions to the microcontroller of the flight control of the unmanned aerial vehicle by using wireless communication, the microcontroller controls the hyperspectral detector, the fluorescence laser radar system and the IMU / GPS system to synchronously collect data; the unmanned aerial vehicle control system synchronously transmits the POS data of the fluorescence laser radar system to the hyperspectral detector, so that the hyperspectral detector and the fluorescence radar system have the same position information and time information when detecting, and the collected data is transmitted back to the ground station computer through wireless connection; wherein, the POS data is the position data and attitude data of the unmanned aerial vehicle.

3. The method according to claim 1, wherein the method is characterized by: In the step one, the resampling method is used to match the spatial resolution of the spectral data obtained by remote sensing detection and the point cloud data generated by the laser radar system, so as to realize the spatial consistency of the hyperspectral data and the fluorescence laser radar point cloud data; If the image resolution of the spectral image data is lower than the spatial resolution of the laser radar point cloud data, the spectral image data is resampled by using the interpolation method; On the contrary, if the image resolution of the spectral image data is higher than the spatial resolution of the laser radar point cloud data, the laser radar point cloud data is resampled by using the up-sampling method. The laser radar point cloud image is converted into a 2D image after resampling to obtain a point cloud grid, the intensity value of each pixel in the point cloud grid corresponds to the average intensity value of each foot point in the point cloud grid, the point cloud grid is taken as a reference image for spectral image data registration, spectral image data matching is performed on spectral image data matched with the waveband of the laser radar detector as a template image, so that the spectral image data and the fluorescent laser radar point cloud data can be spatially corresponding.

4. The method according to claim 1, wherein the method is characterized by: In the step one, the CDOM is detected by the hyperspectral sensor carried by the unmanned aerial vehicle to shoot the hyperspectral image of the investigation point, and the DN value of each waveband of the target position can be directly obtained from the hyperspectral image. The DN value is a value quantified by the hyperspectral sensor receiving radiation, which is related to the quantization depth and has no unit and no actual meaning. The reflectance data is converted by radiation calibration. The specific method of radiation calibration is to place a calibration plate under the lens of the hyperspectral detector, and the ambient light intensity is obtained through the calibration plate to calculate the water surface reflectance.

5. The method according to claim 1, wherein the method is characterized by: In the step one, the process of detecting FDOM in water by the fluorescent laser radar system is as follows: the laser of the fluorescent laser radar system emits laser of a specific wavelength, the laser is reflected by the emission mirror of the system and finally hits the water surface at a specific angle, and the FDOM in the water absorbs the excitation light and emits fluorescence with a longer wavelength than the excitation light; the fluorescent signal emitted by the FDOM is received by the telescope of the fluorescent laser radar system, passes through the collimating lens, passes through the optical filter to the photosensitive sensor, and is received by the system.

6. The method according to claim 1, wherein the method is characterized by: In the step one, the actual sampling depth is 0.5m below the water surface, the brown bottle constant temperature box is used for low temperature and light shielding, and the ultraviolet photometer and the spectrophotometer are used to measure the CDOM concentration and the FDOM concentration of the sample respectively.

7. The method according to claim 1, wherein the method is characterized by: In the step two, the specific operation of establishing the CDOM inversion model is as follows: the original spectral information of the target position with the wavelength of 400-1000nm is extracted from the obtained hyperspectral image, and the original spectral data is converted into reflectance spectral data by radiation calibration. The radiation calibration method is as follows: the hyperspectral reflectance information is obtained by the reflected light intensity and the incident light intensity, the reflected light intensity is the DN value of the current environment, and the incident light intensity is the incident light intensity of the calibration plate. Because the reflectance and the reflected light intensity of the calibration plate are known, the incident light intensity of the current environment can be obtained. The calculation formula of the hyperspectral reflectance is as follows: where: p1 is the reflectance of the calibration panel; DN1 is the DN value of the calibration panel; DN t is the DN value of the survey point; p t is the reflectance of the survey point; In spectral analysis, the intensity value of the spectral data will be different due to the influence of the concentration, temperature and pressure factors of different samples. In order to eliminate the interference of these factors and more accurately analyze the relationship between the measured concentration data and the spectral remote sensing reflectance, the spectral data is normalized, and the normalization formula is as follows: wherein: λ i is the ith wavelength; i has a range of 400-1000; L(λ i ) is the initial value of the reflectivity at wavelength λ i ; n is the number of wavelengths contained in the wavelength range; L N (λ i ) is the wavelength λ i normalized reflectance; The spectral interval is divided into a wave band by the intra-spectral segmentation method, the reflectivity mean value in the wave band is taken as the reflectivity data of the whole wave band, the remote sensing reflectivity data of each wave band and the CDOM concentration are analyzed by Pearson through wave band iteration, the correlation coefficient graph is constructed, the wave band reflectivity ratio factor with high correlation is selected to participate in modeling, the absolute value of the Pearson correlation coefficient is higher, the wave band can better represent the CDOM concentration change, and the expression of the correlation coefficient R is as follows: where: x i is the reflectance of the i-th survey site for a certain band; is the mean reflectance of all survey sites for the same band; y i is the concentration of the water quality parameter of the i-th survey site; is the mean concentration of all survey sites; R is the correlation coefficient value; The reflectivity of the wave band with the highest correlation is selected as the independent variable, and the CDOM concentration value is selected as the dependent variable, and a linear inversion formula for inverting the CDOM concentration is established. C CDOM = k1 x R rs + a, In the formula, R rs is the remote sensing reflectivity data of the highest correlation wave band; k1 is the coefficient of the inversion formula; and a is the intercept of the inversion formula, and the values of k1 and a are determined by combining measured data analysis. The CDOM concentration of all survey points can be inverted by combining the CDOM inversion model and the hyperspectral data of the survey points.

8. The method according to claim 1, wherein the method is characterized by: In step three, the specific operation for establishing the FDOM inversion model is that when the short-wavelength light is used as the excitation wavelength, the FDOM absorbs and releases the fluorescence with a longer wavelength; Fn(355) is defined as the maximum fluorescence intensity of the emission wavelength Em in 440-470 nm when the excitation wavelength Ex is 355 nm, which represents the relative concentration level of the humic-like substance in the FDOM; the humic-like substance is a component with a relatively high content in the FDOM and is used to represent the concentration of the FDOM, and the Fn(355) parameter of the water body is usually used to represent the concentration of the FDOM of the water body; According to the linear relationship between Fn(355) and the FDOM concentration, the fluorescence peak intensity empirical model inversion algorithm formula is as follows: C FDOM = k2 x Fn(355) + β, In the formula, Fn(355) is the maximum fluorescence intensity of the emission wavelength Em in 440-470 nm when the excitation wavelength Ex is 355 nm; k2 is the coefficient of the inversion formula; and β is the intercept of the inversion formula, and the values of k2 and β are determined by combining the measured data analysis. The FDOM concentration of all survey points can be inverted by combining the FDOM inversion model and the fluorescence spectrum data of the survey points.

9. The method according to claim 1, wherein the method is characterized by: In step four, the calculation formula of the NFDOM concentration is as follows: C NFDOM = C CDOM - C FDOM , where: C CDOM is the CDOM concentration data for the survey points; C FDOM is the FDOM concentration data for the survey points; C NFDOM is the NFDOM concentration data for the survey points.

Images