Inland lake chlorophyll inversion method and system based on hyperspectral data

Abstract

The invention relates to the technical field of lake chlorophyll a inversion, and provides an inland lake chlorophyll a inversion method and system based on hyperspectral data, and the content of chlorophyll a in a lake range is inverted through a multiband index model and a regression method. Firstly, hyperspectral image data of a lake is obtained and preprocessed to remove noise and improve data quality. Then, extracting a water body range by adopting an NDWI index, and extracting wave band information of different sampling points; based on optical characteristics of lake water quality, a multiband index model considering pigment absorption and remote sensing reflectivity is constructed, and an inversion model is optimized in combination with regression analysis. According to the method, the hyperspectral remote sensing image data are collected and fused only through actual measurement of a small number of points, the constructed model is applied to the lake chlorophyll a distribution characteristics on the scale of a monitoring area, accurate inversion of the concentration of the chlorophyll a in the whole area of the lake can be achieved, and the problem that the accuracy is low when remote sensing data is singly adopted for inversion is solved.

Description

Technical Field

[0001] The present disclosure relates to the technical field related to the inversion of lake chlorophyll a. Specifically, it relates to a method and system for inverting inland lake chlorophyll a based on hyperspectral data. Background Art

[0002] The statements in this part merely provide background technical information related to the present disclosure and do not necessarily constitute prior art.

[0003] Under the combined influence of global climate change and human activities, lake ecosystems face serious challenges, manifested as shrinking lake areas, increasing eutrophication, frequent algal blooms, deteriorating water quality, and degraded ecosystem functions. Monitoring results based on satellite data show that as of the summer of 2012, among the 2,058 lakes with an area greater than 25 km² globally, more than 63.1% of the lakes were in a highly eutrophic state, 39.4% were moderately eutrophic, and only 10.7% were oligotrophic. Therefore, the monitoring of lake eutrophication has become a key requirement for the current environmental governance of inland lakes. Chlorophyll-a is a common pigment of water algae, an important indicator for measuring the degree of water eutrophication, and also a key parameter in the global water carbon cycle. Obtaining long-term sequence data of lake chlorophyll a content is of great significance for scientific research, social and economic decision-making, and environmental management.

[0004] Traditional water quality monitoring methods mainly rely on field sampling and laboratory chemical analysis to evaluate the water body status through data from isolated sampling points. However, limited by the lake area, sampling density, and experimental costs, this method not only consumes a large amount of manpower, material resources, and financial resources but also has a limited data coverage, making it difficult to comprehensively reflect the spatio-temporal variation characteristics of lake water quality. Compared with traditional monitoring means, remote sensing technology can obtain water quality information more quickly and comprehensively due to its advantages such as wide coverage, stable monitoring cycle, and low cost. In recent years, the monitoring of lake chlorophyll a mainly relies on ocean color satellites (such as MODIS, MERIS, OLCI, etc.) to analyze the changes in lake water quality using long-term time series data. However, the spatial resolution of ocean color remote sensing satellites is relatively low (300 - 1000m), which is suitable for large-area water body monitoring but cannot meet the refined monitoring requirements for small lakes, bays, and small and medium-sized reservoirs. In addition, current remote sensing monitoring mainly relies on spaceborne multi-spectral remote sensing images, but multi-spectral remote sensing has problems such as limited spectral bands, lack of spectral information, and relatively low spatial resolution, making it difficult to meet the requirements for high-precision water quality parameter inversion. At the same time, spaceborne remote sensing is greatly affected by atmospheric conditions, and the data is easily interfered by clouds and fog, further restricting the monitoring accuracy and data availability. Currently, the water quality monitoring field mainly relies on spaceborne multi-spectral remote sensing images as data sources. However, such images have some limitations: First, the spectral bands are few and discontinuous, making it difficult to accurately identify the spectral characteristics of inland water bodies; second, the spatial resolution is relatively coarse, making it difficult to meet the requirements for refined remote sensing quantitative inversion. In addition, the acquisition cycle of spaceborne remote sensing data is long, and it is easily affected by atmospheric conditions, and there are often a large number of clouds and fog in the generated images, which are difficult to effectively remove. At the same time, inland water bodies are usually small in area and fragmented, with complex optical characteristics, and are easily affected by weather, the surrounding environment, and human activities. These factors further reduce the accuracy of water quality parameter inversion, resulting in a large error between the inversion results and the actual targets. There are various difficulties in inverting the water quality of water bodies.

[0005] In the remote sensing monitoring of lake water quality parameters, the inversion of chlorophyll a currently mainly uses three models: empirical method, analytical method, and semi-analytical method. Among them, the empirical method is applicable to clean water bodies. However, for turbid lakes (class II water bodies), their optical characteristics are complex and are affected by multiple factors such as chlorophyll a, total suspended solids, and yellow substances. Simply using the empirical method is difficult to ensure universality. In addition, the empirical method usually relies on data from specific years to establish models and lacks applicability across time and regions. Although the analytical method can provide higher-precision inversion results, it requires a large amount of on-site synchronous observation data, which is difficult to obtain. Therefore, the semi-analytical method has become the current mainstream method and has certain advantages in the quantitative inversion of water quality parameters. However, the complexity of lake water quality still leads to limitations of these methods in practical applications.

[0006] In summary, a single remote sensing data cannot accurately retrieve chlorophyll a. The existing chlorophyll a retrieval method mainly targets marine water types. Due to the particularity of the lake environment, the water body is affected by multiple factors, such as suspended solids, phytoplankton, dissolved organic matter, etc. The combined effect of these factors results in complex optical properties. At the same time, inland lakes often have high turbidity, which increases the scattering and absorption of light, reduces the quality of remote sensing signals, and makes it difficult to retrieve water quality parameters solely based on remote sensing signals. When applied to lake waters, the retrieval effect is often poor; moreover, the complexity of lake water quality leads to limitations of existing methods in practical applications. Summary of the Invention

[0007] To solve the above problems, the present disclosure proposes a method and system for retrieving chlorophyll a in inland lakes based on hyperspectral data. For the method of retrieving chlorophyll a in inland lakes, this method can realize the inversion analysis of the water body in complex inland lakes and can fully reflect the spatial distribution characteristics of water quality parameters in the whole lake; it can enhance the applicable range and accuracy of the retrieval method, provide a scientific basis for inland lake water quality monitoring, and also provide theoretical support and reference for the treatment and improvement of inland lake water quality.

[0008] To achieve the above object, the present disclosure adopts the following technical solutions:

[0009] One or more embodiments provide a method for retrieving chlorophyll a in inland lakes based on hyperspectral data, including the following steps:

[0010] Collect lake water samples in the field to construct a measured data set of chlorophyll a in the lake;

[0011] Obtain a data set of remote sensing reflectance of the lake, and extract hyperspectral satellite image data at the measured sampling time for preprocessing;

[0012] For the preprocessed hyperspectral satellite image data, extract the water body range through the NDWI index operation, extract band information, and obtain band data corresponding to each measured point and a remote sensing image of the water area of the lake range;

[0013] Construct a multi-band index model considering the influence of pigment absorption and remote sensing reflectance, fuse the multi-band index model and the regression model to construct an inversion model, use the data of the measured data set as the output, and use the band data of each measured point and the remote sensing image of the water area of the lake range as the input, and iteratively optimize the constructed inversion model;

[0014] Obtain hyperspectral satellite image data at the position to be measured, and input it into the optimized inversion model to obtain the inversion result of chlorophyll a at the position to be measured.

[0015] One or more embodiments provide a system for retrieving chlorophyll a in inland lakes based on hyperspectral data, including:

[0016] An in-situ measurement module, configured to construct an in-situ measurement dataset of lake chlorophyll a by collecting lake water samples in the field;

[0017] A first remote sensing data processing module, configured to obtain a dataset of lake remote sensing reflectance, extract hyperspectral satellite image data at the in-situ measurement time for preprocessing;

[0018] A second remote sensing data processing module, configured to extract the water body range by NDWI index operation from the preprocessed hyperspectral satellite image data, perform band information extraction, and obtain band data corresponding to each in-situ measurement point and a remote sensing image of the water area of the lake range;

[0019] A model optimization module, configured to construct a multi-band index model considering the influence of pigment absorption and remote sensing reflectance, fuse the multi-band index model and a regression model to construct an inversion model, and perform iterative optimization on the constructed inversion model with the data of the in-situ measurement dataset as the output and the band data corresponding to each in-situ measurement point and the remote sensing image of the water area of the lake range as the input;

[0020] An inversion module, configured to obtain hyperspectral satellite image data at a position to be measured and input it into the optimized inversion model to obtain an inversion result of chlorophyll a at the position to be measured.

[0021] An electronic device, including a memory, a processor, and computer instructions stored on the memory and running on the processor. When the computer instructions are run by the processor, the steps in the above-mentioned method for inverting chlorophyll a in inland lakes based on hyperspectral data are completed.

[0022] A computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the steps in the above-mentioned method for inverting chlorophyll a in inland lakes based on hyperspectral data are completed.

[0023] Compared with the prior art, the beneficial effects of the present disclosure are:

[0024] The method disclosed herein utilizes hyperspectral satellite image data in combination with measured data to retrieve the chlorophyll-a content within the lake area through multi-band index models and regression methods. First, hyperspectral image data of the lake is acquired and preprocessed to remove noise and improve data quality. Then, the Normalized Difference Water Index (NDWI) is used to extract the water body area, and the band information at different sampling points is extracted. Based on the optical properties of lake water quality, a multi-band index model considering pigment absorption and remote sensing reflectance is constructed, and the inversion model is optimized through regression analysis. Finally, the optimized model is used to process the hyperspectral images of the area to be measured; through the above method, only the measured data at a small number of points is collected to fuse the hyperspectral remote sensing image data, and a model constructed by combining bi-optical principles and statistical methods is applied to monitor the distribution characteristics of chlorophyll-a in the lake at the regional scale, achieving the accurate inversion of the chlorophyll-a concentration in the entire lake area, realizing the chlorophyll-a water quality detection of the entire lake area, integrating regional and macroscopic aspects, and solving the problem of low accuracy in inversion using only remote sensing data alone.

[0025] The advantages of the present disclosure and the advantages of additional aspects will be described in detail in the following specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The accompanying drawings forming a part of this disclosure are used to provide a further understanding of the present disclosure. The schematic embodiments and descriptions thereof of the present disclosure are used to explain the present disclosure and do not constitute a limitation to the present disclosure.

[0027] Figure 1 is a flowchart of the inversion method of Embodiment 1 of the present disclosure;

[0028] Figure 2 is a schematic diagram of the accuracy evaluation results of the chlorophyll inversion model based on the hyperspectral data of ZY-1 02E satellite in the experimental example of Embodiment 1 of the present disclosure;

[0029] Figure 3 is a chlorophyll inversion result map based on the hyperspectral data of ZY-1 02E satellite in the experimental example of Embodiment 1 of the present disclosure; DETAILED DESCRIPTION OF THE EMBODIMENTS

[0030] The present disclosure will be further described below in conjunction with the accompanying drawings and embodiments.

[0031] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present disclosure belongs.

[0032] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present disclosure. As used herein, unless the context clearly indicates otherwise, the singular forms are also intended to include the plural forms. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they specify the presence of features, steps, operations, devices, components, and / or combinations thereof. It should be noted that, without conflict, the various embodiments and features in the present disclosure can be combined with each other. The embodiments will be described in detail below with reference to the drawings.

[0033] Embodiment 1

[0034] In the technical solutions disclosed in one or more embodiments, as Figures 1 to 3 shown, the method for inverting chlorophyll a in inland lakes based on hyperspectral data includes the following steps:

[0035] Step 1: Construct a measured dataset of lake chlorophyll a by collecting lake water samples in the field and obtain a dataset of remote sensing reflectance of the lake;

[0036] Step 2: Extract hyperspectral satellite image data at the measured sampling time from the dataset of remote sensing reflectance of the lake for preprocessing;

[0037] Step 3: Use the NDWI index operation to extract the water body range from the preprocessed hyperspectral satellite image data, extract band information, and obtain band data corresponding to each measured point and a remote sensing image of the water area within the lake range;

[0038] Step 4: Construct a multi-band index model considering the influence of pigment absorption and remote sensing reflectance, fuse the multi-band index model and the regression model to construct an inversion model, use the data of the measured dataset as the output, and use the band data corresponding to each measured point and the remote sensing image of the water area within the lake range as the input, and iteratively optimize the constructed inversion model;

[0039] Step 5: Obtain hyperspectral satellite image data at the position to be measured, and input it into the optimized model to obtain the inversion result of chlorophyll a at the position to be measured.

[0040] In this embodiment, the method uses hyperspectral satellite image data combined with measured data to retrieve the chlorophyll-a content within the lake range through a multi-band index model and regression method. First, hyperspectral image data of the lake is obtained and preprocessed to remove noise and improve data quality. Then, the NDWI index is used to extract the water body range, and the band information of different sampling points is extracted. Based on the optical characteristics of lake water quality, a multi-band index model considering pigment absorption and remote sensing reflectance is constructed, and the inversion model is optimized by combining regression analysis. Finally, the optimized model is used to process the hyperspectral images of the area to be measured. Through the above method, by only collecting and fusing the hyperspectral remote sensing image data at a small number of points through actual measurement, and further applying the constructed model to monitor the distribution characteristics of chlorophyll-a in the lake at the regional scale, the accurate inversion of the chlorophyll-a concentration in the entire lake area can be achieved, realizing the chlorophyll-a water quality detection of the entire lake area, integrating regional and macroscopic aspects, and solving the problem of low accuracy in inversion using only remote sensing data.

[0041] The algorithm flow of this embodiment has higher applicability, can ensure better results after being applied to complex water bodies of inland lakes, and complete the detection of the spatio-temporal dynamic changes of the water quality of the entire lake surface in different lakes. This applicability makes the present invention not only applicable to the water area type of inland high-turbidity lakes, but also has high accuracy in water quality monitoring, which is of great significance for the future development of satellite data monitoring.

[0042] The above method first conducts on-site multi-point sampling for the area to be detected (such as a certain lake), optimizes the trained inversion model based on the measured data from the sampling and the data points corresponding to the positions to be detected in the satellite remote sensing data, and finally conducts a comprehensive detection and analysis of the area to be detected based on the model optimized using the data of the area to be detected. When replacing the lake to be detected, the same method is used to optimize and re-detect the model.

[0043] In step 1, by collecting lake water samples in the field, a lake chlorophyll-a dataset and a lake remote sensing reflectance dataset are obtained. The lake chlorophyll-a dataset is used as the output of the inversion model;

[0044] The measured dataset of lake chlorophyll-a includes the coordinate information of the measured positions of lake chlorophyll-a, the concentration of chlorophyll-a measured at the corresponding positions, and spectral data;

[0045] The coordinate information is provided for step 2 to obtain the image data at the corresponding positions as the input of the inversion model, the chlorophyll-a concentration is used as the output of the inversion model, and the spectral data provides a reference for the iteration range of the inversion model.

[0046] The lake remote sensing reflectance dataset is the image data obtained by hyperspectral satellites, including data such as reflectance corresponding to the position coordinates; the remote sensing image data is the hyperspectral satellite image data. In this embodiment, the hyperspectral satellite image of Lake Resource No. 1 is collected.

[0047] In this step, the obtained remote sensing image data corresponds to the measured sampling time in Step 1 and is the sampling data within the same time period.

[0048] Furthermore, the method for preprocessing the obtained remote sensing image data includes the following methods:

[0049] 1) Perform radiometric calibration processing, and convert the DN value into a radiance image based on the radiometric calibration parameters. The calculation formula is:

[0050] L = G·DN + B;

[0051] where L is the radiance; DN is the original digital quantization value; G is the gain coefficient (Gain); B is the offset (Bias);

[0052] 2) Convert the data after radiometric calibration processing into the BIL storage format and perform atmospheric correction.

[0053] Specifically, for ZY-1 02E satellite data, the sensor altitude is set to 778 km, the water vapor inversion selects the 940 nm band, and the number of bands for spectral polishing is set to 3. The FLAASH atmospheric correction algorithm can be used for atmospheric correction.

[0054] In this embodiment, different image processing methods are adopted for the remote sensing data, and preprocessing such as radiometric calibration, atmospheric correction, and geometric correction is performed on the remote sensing images, which can ensure the information extraction accuracy of each type of data and extract the details of the image features, thereby improving the accuracy of subsequent chlorophyll a inversion.

[0055] In Step 3, for the preprocessed remote sensing image data, extract the water body range and band information to obtain the remote sensing images of each band data and the water area of the lake to be measured in the remote sensing image data.

[0056] Step 31 Water body range extraction:

[0057] 31.1) For the obtained remote sensing image data, calculate the NDWI index. The calculation formula is as follows:

[0058]

[0059] where Green represents the reflectance of the green band in the remote sensing image data, and NIR is the reflectance of the red band in the remote sensing image data;

[0060] 31.2) Set the threshold to zero. By comparing the NDWI index with the threshold, the area greater than the threshold is designated as the water area;

[0061] In this embodiment, when NDWI > 0, it indicates that the area is water. Areas with higher NDWI values usually represent water, while vegetation, soil, etc. are non-water areas.

[0062] 31.3) Crop the remote sensing image, retain the water area, and remove the non-water area.

[0063] In the above solution, by calculating the normalized difference between the green band and the near-infrared band to enhance water information and at the same time suppress the influence of vegetation and soil, high-precision extraction of the water range can be achieved;

[0064] In step 32, band information is extracted according to the measured points of lake chlorophyll a data as the input of the inversion model; that is, according to the measured points of lake chlorophyll a data, the band information at the corresponding measured point positions in the remote sensing image is extracted.

[0065] 32.1) For the remote sensing image data of the extracted water range, perform image reprojection to match the coordinate system of the measured point data so that the coordinate system of the remote sensing image is consistent with the measured points;

[0066] 32.2) Obtain the measured points, and extract the band raster values of the remote sensing image at the corresponding measured point coordinate positions as the extracted band information;

[0067] In this embodiment, for the extraction of band information, according to the coordinate positions of the measured lake chlorophyll a samples, the raster value of each band of the image corresponding to the samples is extracted as the remote sensing reflectance corresponding to the measured lake samples;

[0068] Among them, the raster value is the value of the remote sensing image pixel (picture element), which usually represents the reflectance or radiance of the pixel in a certain band; in remote sensing data processing, each pixel has corresponding geographical coordinates (latitude and longitude or projected coordinates) and values of multiple bands, and these values are used to analyze surface features.

[0069] In step 4, for the constructed inversion model, combining the measured spectral data and the iterative algorithm, determine the parameter settings of the multi-band index model, and screen through various regression models to determine the optimal inversion model;

[0070] Optionally, the multi-band index model adopts a three-band index model or a four-band index model. Preferably, in this embodiment, a three-band index model is adopted;

[0071] Furthermore, the method for constructing the inversion model includes the following steps:

[0072] Step 41: Considering the influence of pigment absorption and remote sensing reflectance, select the optimal bands.

[0073] Step 42: Based on the selected optimal bands, construct a multi-band index model and calculate the three-band index.

[0074] In this embodiment, the specific form of the three-band index model constructed considering the influence of pigment absorption and remote sensing reflectance is:

[0075] x = (R rs (λ1) -1 -R rs (λ2) -1 ) * R rs (λ3)

[0076] where λ1 represents the first optimal band, λ2 represents the second optimal band, λ3 represents the third optimal band; R rs represents the reflectance of the corresponding band.

[0077] When constructing the three-band index model, the method for selecting the optimal bands includes the following steps:

[0078] 4.1) The first optimal band λ1 is selected as the chlorophyll a absorption peak band.

[0079] Specifically, the first optimal band λ1 is selected at the position where the influence of pigment absorption on remote sensing reflectance is the greatest, and the influence of the absorption of xanthophyll and non-pigment particles on the total backscattering on remote sensing reflectance is relatively small, that is, at the chlorophyll a absorption peak within the red light band.

[0080] 4.2) The second optimal band λ2 is selected as the chlorophyll a fluorescence peak band.

[0081] Specifically, the second optimal band λ2 is selected near λ1 and at a position where the absorption of chlorophyll a is relatively small, that is, the chlorophyll a fluorescence peak within the red light band can meet the selection requirements of λ2.

[0082] 4.3) The third optimal band λ3 is selected as the pure water absorption band.

[0083] Specifically, the selection condition for the third optimal band λ3 is: the band where the total absorption coefficient is much greater than the backscattering coefficient and is not affected by the absorption of chlorophyll, non-pigment particles, and xanthophyll, that is, the remote sensing reflectance is mainly affected by the absorption of pure water.

[0084] In this embodiment, innovative selection conditions are proposed to select three optimal bands and construct an index to improve the inversion accuracy, which is particularly suitable for water quality monitoring in highly turbid waters.

[0085] Further, after selecting the optimal bands of the three-band index model, the determination of the three optimal bands of λ1, λ2, and λ3 uses the characteristic band range for iterative optimization as the final optimal bands, that is, through iteration, the correlation analysis with the lake chlorophyll a concentration is carried out respectively, including the following steps:

[0086] Step 4-1: According to the spectral data obtained from the field collection of lake water samples in Step 1, initially determine the characteristic band range for the determination of the three optimal bands as the initial band ranges of λ1, λ2, and λ3;

[0087] For example, in this embodiment, for the spectral data of a certain lake, the characteristic band range is used to determine the three optimal bands of λ1, λ2, and λ3 according to the measured spectral curve, where the value range of λ1 is 650 - 680 nm, the value range of λ2 is 680 - 720 nm, and the value range of λ3 is 700 - 770 nm;

[0088] Step 4-2: Based on the initial band ranges corresponding to the corresponding satellites, determine the band ranges corresponding to the corresponding satellites, and broaden the band selection ranges determined for the second optimal band λ2 and the third optimal band λ3 based on turbidity to obtain the band selection ranges of λ1, λ2, and λ3;

[0089] According to the value ranges of the three bands, the equivalent value ranges for the hyperspectral data of ZY-1 02E satellite are: the value range of λ1 is B32 - B36, the value range of λ2 is B36 - B42, and the value range of λ3 is B38 - B46. Iterations are carried out respectively within the three band selection ranges;

[0090] Step 4-3: In the band selection ranges of λ1, λ2, and λ3, fix any two bands, and iterate the remaining one band within the band selection range. Each time, calculate the correlation between the corresponding band data and the concentration of chlorophyll a, and identify the one with the highest correlation as the output band to obtain the optimal bands of the three bands;

[0091] Specifically, starting from λ1, fix the λ2 and λ3 bands as B39 and B42, and iterate within the iteration range of λ1. The band with the largest correlation in the results is used as the output band of λ1. Then, perform the iteration of λ2. After completing the iteration of λ2, continue to iterate λ3 until the iteration of the three bands is completed. Iterate λ1 again. If the result band is the same as the first iteration, the three selected bands are the optimal bands; otherwise, continue to iterate until the same iteration result appears.

[0092] Optionally, the Pearson simple correlation coefficient is used as the reference index in the iteration process, and the root mean square error, etc. can also be selected as the iteration index.

[0093] Further, in step 4-2, based on the turbidity of the lake area obtained from on-site sampling, the band selection range determined by the second optimal band λ2 and the third optimal band λ3 among the three optimal bands is widened as the selected band selection range;

[0094] Specifically, the formula for widening the band selection range is:

[0095]

[0096] Wherein, is the maximum value of the suspended sediment concentration in season t of the lake within the set time period, is the minimum value of the suspended sediment concentration in season t of the lake within the set time period, C is the average value of the suspended sediment concentration in this water area, and E is the widening ratio;

[0097] Optionally, the set time period can be any time period such as several years or several months. For example, if the set time is three years, the suspended sediment concentration corresponding to season t of each year can be taken, and the maximum value within three years can be found as The minimum value is taken as

[0098] For high-turbidity water areas, due to the high concentration of suspended sediments, the response of remote sensing reflectance to chlorophyll a concentration is less obvious. Therefore, the characteristic band range of the three optimal bands of the three-band index for high-turbidity water areas should be widened and improved;

[0099] In this embodiment, through the widening of the band range, as the water turbidity increases, the widening degree of the band selection range can achieve full coverage. The widened band range can cover more spectral features and capture the reflectance changes caused by suspended sediments, thereby improving the sensitivity of remote sensing data to suspended sediments. The band selection range widened by the widening formula of this embodiment can reduce the instability of the inversion model caused by insufficient spectral change range. Especially in high-turbidity areas, the inversion model can more stably extract water quality information, reduce outliers and deviations, and enhance the prediction ability of the model.

[0100] In the above solution, by constructing a three-band index model, according to the measured spectral morphological characteristics, a suitable band range is selected, and the correlation with the measured chlorophyll a concentration is calculated respectively through an iterative algorithm. The above process constructs a three-band index model. Next, a chlorophyll a remote sensing inversion model suitable for lakes is established through regression model fusion processing.

[0101] Step 43: Using the three-band index calculated by the multi-band index model as the independent variable and the concentration of chlorophyll a in the measured dataset of measured chlorophyll a as the output, establish multiple regression models for fitting;

[0102] Due to the influence of high suspended solids in the lake, the relationship between chlorophyll a content and the three-band index is sometimes not only linear. Different regions or different satellite data may adopt different regression models after screening. Therefore, five regression models, namely linear, quadratic polynomial, cubic polynomial, exponential, and power models, are established to determine a closer relationship between the two. The specific forms are as follows:

[0103] y = ax + b;

[0104] y = ax 2 + bx + c;

[0105] y = ax 3 + bx 2 + cx + d;

[0106] y = ae bx ;

[0107] y = (ax) b ;

[0108] Among them, y is the chlorophyll a concentration, x is the three-band index calculated by the multi-band index model, and a, b, and c are coefficients that can be obtained after multiple fittings;

[0109] Step 44: Set evaluation indicators, perform leave-one-out cross-validation and error evaluation on the fitting results, select the optimal regression model, and fuse it with the constructed multi-band index model as the final inversion model;

[0110] The fitting accuracy of the leave-one-out cross-validation LOOCV (Leave-One-Out Cross Validation) regression model is adopted. Specifically: One sample out of N samples is used as the independent validation set, and the remaining N - 1 samples are used for modeling, and then it loops N times. Then, accuracy tests are carried out to test the accuracy and stability of the model for later selection of a model with better accuracy and higher stability.

[0111] The evaluation indicators constructed for error evaluation can include: coefficient of determination (i.e., R2), root mean square error (i.e., RMSE), mean absolute percentage error (i.e., MAPE). The larger the R2, the smaller the RMSE, and the smaller the MAPE, the higher the model accuracy and the better the stability; Select the regression model with the highest accuracy and best stability as the final model, and fuse it with the multi-band index model to form the inversion model.

[0112] In step 5, the hyperspectral satellite image data to be measured is obtained. First, it undergoes the preprocessing in step 2. After calculating the three-band index through the multi-band index model in the inversion model, it is used as the input of the regression model, and the inversion calculation is carried out through the optimal regression model selected in the inversion model to obtain the concentration of chlorophyll a.

[0113] In this embodiment, by combining the lake water sample data collected on-site with the hyperspectral remote sensing image data, a method for remotely retrieving chlorophyll-a in lakes based on remote sensing images is proposed. First, a measured data set of lake chlorophyll-a is obtained through on-site water sample collection, providing reliable reference data for subsequent inversion models. Second, the hyperspectral satellite image of Lake Resources No. 1 is used, and preprocessing such as radiometric calibration, atmospheric correction, and geometric correction of the image is performed through ENVI software to ensure that the accuracy and details of the image data are retained. On this basis, the water body range is extracted through NDWI index operation, and further band information extraction is carried out to accurately obtain the image data related to the chlorophyll-a concentration. Finally, a three-band index model is constructed, and the correlation between band selection and chlorophyll-a concentration is optimized through an iterative algorithm, thereby realizing the construction of a high-precision remote sensing inversion model for lake chlorophyll-a. This technical solution can effectively improve the accuracy of chlorophyll-a concentration inversion and provide a scientific basis for lake ecological monitoring and water quality evaluation.

[0114] To illustrate the effect of the above method, the inversion method of this embodiment is applied to the ZY-1 02E hyperspectral remote sensing data for verification of technical effects;

[0115] (1) The experimental area of this example is a certain lake. A synchronous experiment with the ZY-1 02E (Resources No. 1 02E satellite) satellite is carried out in the water area of the experimental lake. Water surface sampling points are arranged, and spectral measurement and determination of chlorophyll-a in surface water samples are completed on-site and the position information is recorded. A total of 20 sample points are collected.

[0116] (2) Obtain the ZY-1 02E hyperspectral data on the experimental day. The image quality is good and there are no clouds blocking above the water body in the study area. Based on the radiometric calibration parameters, the DN value is converted into a radiance image and converted into the BIL storage format, and then the FLAASH module of ENVI is used to perform atmospheric correction, and finally geometric correction of the image is carried out.

[0117] (3) Use ENVI software to calculate the NDWI index for the remote sensing image. Set the threshold that when NDWI>0, it indicates that the area is a water body, and the remote sensing image is cropped. The calculation formula of NDWI for the ZY-1 02E satellite's bands is as follows:

[0118]

[0119] Synchronously, according to the position information recorded during the synchronous experiment in the water area of the experimental lake, band information extraction is carried out on the processed remote sensing image data;

[0120] (4) Construct a three - band index model. Determine the three optimal bands λ1, λ2, and λ3 according to the measured spectral curves using the characteristic band ranges. The value range of λ1 is 650 - 680 nm, the value range of λ2 is 680 - 720 nm, and the value range of λ3 is 700 - 770 nm. The equivalent value ranges for the hyperspectral data of ZY - 1 02E satellite are: the value range of λ1 is B32 - B36, the value range of λ2 is B36 - B42, and the value range of λ3 is B38 - B46. Conduct iterations within the selection ranges of the three bands respectively; the final iteration results are λ1 is B34, λ2 is B39, and λ3 is B42.

[0121] The correlation is calculated using the Pearson simple correlation coefficient method. The larger the absolute value of the correlation coefficient, the stronger the correlation. The closer the correlation coefficient is to 1 or - 1, the stronger the correlation degree; the closer the correlation coefficient is to 0, the weaker the correlation degree. Using the three - band model obtained from the above iteration as the independent variable and the chlorophyll a content as the dependent variable, the regression fitting method is used. Due to the influence of high suspended matter in the lake, the relationship between the chlorophyll a content and the three - band index is sometimes not just linear. Therefore, five regression models, namely linear, quadratic polynomial, cubic polynomial, exponential, and power models, are established to explore a closer relationship between the two. The chlorophyll a inversion model is as follows:

[0122] Chla=92.97*(R rs (B34) -1 -R rs (B39) -1 )*R rs (B42)+10.18

[0123] (4) Apply the established optimal chlorophyll a remote - sensing inversion model to the domestic Resource Satellite No.1 hyperspectral remote - sensing images of the study area, and obtain the remote - sensing inversion map of chlorophyll a content in the lake water area as Figure 2 shown. It can be seen from Figure 2 that there is a high degree of fitting between the calculated value of chlorophyll a inverted by the model and the measured value. The determination coefficient R2 = 0.91 indicates that the model has a strong explanatory ability. The root - mean - square error (RMSE) is 1.01, and the mean absolute percentage error (MAPE) is 8.52%, indicating that the error of the model in estimating the chlorophyll a concentration is small and it has high precision. In addition, most of the data points are distributed near the 1:1 reference line, indicating that the model has good reliability in inverting the chlorophyll a content in the lake water area and can be used for regional water quality monitoring and ecological environment assessment.

[0124] Specifically, the inversion results are as Figure 3As shown, it can be seen from the figure that after optimizing the inversion model through sampling at multiple measured points, the water quality of the entire lake water area can be accurately inversely calculated.

[0125] Embodiment 2

[0126] Based on Embodiment 1, an inland lake chlorophyll a inversion system based on hyperspectral data is provided in this embodiment, including:

[0127] A measured data module, configured to construct a measured data set of lake chlorophyll a by collecting lake water samples in the field;

[0128] A first remote sensing data processing module, configured to obtain a lake remote sensing reflectance data set, extract hyperspectral satellite image data at the measured sampling time for preprocessing;

[0129] A second remote sensing data processing module, configured to extract the water body range through NDWI index operation on the preprocessed hyperspectral satellite image data, extract band information, and obtain band data corresponding to each measured point and a remote sensing image of the lake range water area;

[0130] A model optimization module, configured to construct a multi-band index model considering the influence of pigment absorption and remote sensing reflectance, fuse the multi-band index model and the regression model to construct an inversion model, use the data of the measured data set as the output, use the band data of each measured point and the remote sensing image of the lake range water area as the input, and iteratively optimize the constructed inversion model;

[0131] An inversion module, configured to obtain hyperspectral satellite image data of the position to be measured, and input it into the optimized inversion model to obtain the inversion result of chlorophyll a at the position to be measured.

[0132] It should be noted here that each module in this embodiment corresponds to each step in Embodiment 1, and the specific implementation process is the same, so it will not be repeated here.

[0133] Embodiment 3

[0134] This embodiment provides an electronic device, including a memory and a processor, and computer instructions stored on the memory and running on the processor. When the computer instructions are run by the processor, the steps in the method for inverting chlorophyll a in an inland lake based on hyperspectral data in Embodiment 1 are completed.

[0135] Embodiment 4

[0136] This embodiment provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the steps in the method for inverting chlorophyll a in an inland lake based on hyperspectral data in Embodiment 1 are completed.

[0137] The above are only the preferred embodiments of the present disclosure and are not intended to limit the present disclosure. For those skilled in the art, the present disclosure may have various changes and modifications. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present disclosure shall be included within the protection scope of the present disclosure.

[0138] Although the specific implementation manners of the present disclosure have been described above in conjunction with the accompanying drawings, it is not a limitation on the protection scope of the present disclosure. Those skilled in the art should understand that, based on the technical solutions of the present disclosure, various modifications or deformations that can be made by those skilled in the art without creative efforts are still within the protection scope of the present disclosure.

Claims

1. The inversion method of chlorophyll a in inland lakes based on hyperspectral data is characterized by: The steps include: By collecting lake water samples on site, a measured data set of lake chlorophyll a was constructed; Obtain lake remote sensing reflectance datasets and extract hyperspectral satellite image data of measured sampling time for preprocessing; The pre-processed hyperspectral satellite image data is used to extract the water body range through NDWI index calculation, and the band information is extracted to obtain the band data corresponding to each measured point and the remote sensing image of the lake range water area; A multi-band index model was constructed by considering the influence of pigment absorption and remote sensing reflectance. The inversion model was constructed by integrating the multi-band index model and the regression model. The data of the measured data set was used as output, and the band data of each measured point and the remote sensing image of the lake area were used as input to iteratively optimize the constructed inversion model. The hyperspectral satellite image data of the location to be measured is obtained and input into the optimized inversion model to obtain the inversion result of chlorophyll a at the location to be measured.

2. The inland lake chlorophyll a inversion method based on hyperspectral data as claimed in claim 1, characterized in that: For the pre-processed remote sensing image data, water body range extraction is performed, including: Calculate the NDWI index based on the remote sensing image data obtained; The threshold is set to zero, and by comparing the NDWI index with the threshold, the area greater than the threshold is defined as the water area; The remote sensing images are cropped to retain the water area and remove the non-water area.

3. The inland lake chlorophyll a inversion method based on hyperspectral data as claimed in claim 1, characterized in that: Band information is extracted based on the measured points of lake chlorophyll a data, including: Reproject the remote sensing image data of the extracted water body range to match the coordinate system of the measured point data, so that the coordinate system of the remote sensing image is consistent with the measured point; Obtain the measured points, and extract the band raster values ​​of the remote sensing image corresponding to the coordinate positions of the measured points as the extracted band information.

4. The inland lake chlorophyll a inversion method based on hyperspectral data as claimed in claim 1, characterized in that: The multi-band index model adopts a three-band index model or a four-band index model.

5. The inland lake chlorophyll a inversion method based on hyperspectral data as claimed in claim 1, characterized in that: The method for constructing the inversion model includes the following steps: Considering the influence of pigment absorption and remote sensing reflectance, the optimal band is selected; Construct a multi-band index model based on the selected optimal bands and calculate the three-band index; Using the three-band index calculated by the multi-band index model as the independent variable and the concentration of chlorophyll a in the measured data set of chlorophyll a as the output, various regression models were established for fitting; Evaluation indicators were set, leave-one-out cross-validation and error evaluation were performed on the fitting results, and the optimal regression model was selected and combined with the constructed multi-band index model as the final inversion model.

6. The inland lake chlorophyll a inversion method based on hyperspectral data as claimed in claim 5, characterized in that: The specific form of constructing a three-band index model considering the influence of pigment absorption and remote sensing reflectance is: x=(R rs (λ1) -1 -R rs (λ2) -1 )*R rs (λ3) Among them, λ1 represents the first optimal band, λ2 represents the second optimal band, and λ3 represents the third optimal band; R rs Indicates the reflectivity of the corresponding band; Alternatively, for the three-band index model, the method for selecting the optimal band includes the following steps: The first optimal band λ1 is selected as the chlorophyll a absorption peak band; The second optimal wavelength λ2 is selected as the chlorophyll a fluorescence peak wavelength; The third optimal band λ3 is selected as the pure water absorption band; Alternatively, the multiple regression models include linear, polynomial, exponential, and / or power regression models.

7. The inland lake chlorophyll a inversion method based on hyperspectral data according to claim 6, characterized in that: After selecting the optimal band of the three-band index model, the determination of the three optimal bands of λ1, λ2 and λ3 is iteratively optimized using the characteristic band range as the final optimal band, including the following steps: Based on the preliminary spectral data obtained from lake water samples collected on site, the preliminary band ranges of λ1, λ2 and λ3 were determined; Based on the preliminary band range corresponding to the corresponding satellite, the band range corresponding to the corresponding satellite is determined, and the band selection range determined by the second optimal band λ2 and the third optimal band λ3 is widened based on the turbidity to obtain the band selection ranges of λ1, λ2 and λ3; In the band selection range of λ1, λ2 and λ3, any two bands are fixed, and the remaining band is iterated within the band selection range. The correlation between the corresponding band data and the concentration of chlorophyll a is calculated in each iteration, and the band with the highest correlation is identified as the output band to obtain the final optimal band of the three bands.

8. The inland lake chlorophyll a inversion system based on hyperspectral data is characterized by: include: The measurement module is configured to construct a measured data set of lake chlorophyll a by collecting lake water samples on site; The first remote sensing data processing module is configured to obtain a lake remote sensing reflectance dataset and extract the hyperspectral satellite image data of the measured sampling time for preprocessing; The second remote sensing data processing module is configured to extract the water range of the pre-processed hyperspectral satellite image data through NDWI index calculation, extract band information, and obtain the band data corresponding to each measured point and the remote sensing image of the lake range water area; The model optimization module is configured to consider the influence of pigment absorption and remote sensing reflectance to construct a multi-band index model, fuse the multi-band index model and the regression model to construct an inversion model, use the data of the measured data set as output, and use the band data of each measured point and the remote sensing image of the lake area as input to iteratively optimize the constructed inversion model; The inversion module is configured to obtain the hyperspectral satellite image data of the position to be measured, and input it into the optimized inversion model to obtain the inversion result of chlorophyll a at the position to be measured.

9. An electronic device, characterized in that: The method comprises a memory and a processor and computer instructions stored in the memory and executed on the processor. When the computer instructions are executed by the processor, the steps in the inversion method for chlorophyll a of inland lakes based on hyperspectral data as described in any one of claims 1 to 7 are completed.

10. A computer-readable storage medium, characterized in that: Used to store computer instructions, which, when executed by a processor, complete the steps of the inversion method for inland lake chlorophyll a based on hyperspectral data as described in any one of claims 1 to 7.

Images