A water body algal toxin concentration estimation method and system based on remote sensing images

Abstract

A water body algae toxin concentration estimation method and system based on remote sensing image belong to the technical field of inland water body eutrophication lake water environment evaluation, solve the technical problems that the existing technology has high precision requirement of detection instrument, expensive equipment material, complex and low efficiency of water sample pretreatment process, and the remote sensing indirect inversion method has poor accuracy and is difficult to realize accurate quantification of microcystin. Step 1, obtain the remote sensing image of the water body, and pretreat to obtain the remote sensing reflectivity data; step 2, calculate the spectral index based on the remote sensing reflectivity data; step 3, construct a water body microcystin concentration remote sensing estimation model, take the spectral index as the input of the water body microcystin concentration remote sensing estimation model, and obtain the water body microcystin concentration. The present application is used for realizing efficient and accurate water body algae toxin concentration estimation.

Description

Technical Field

[0001] This invention relates to the field of water environment assessment technology for eutrophic lakes in inland water bodies, specifically to a method and system for estimating the concentration of algal toxins in water bodies based on remote sensing images. Background Technology

[0002] Microcystin (MC) is a type of algal toxin, a class of biologically active secondary metabolites produced by the metabolism of Microcystis algae in water.

[0003] Currently, the commonly used methods for detecting microcystin both domestically and internationally include the following analytical methods: bioassay, high-performance liquid chromatography (HPLC), enzyme-linked immunosorbent assay (ELISA), mass spectrometry (MS / MS), and sensor detection. Bioassay was the earliest method to emerge, but it has been gradually replaced due to its long detection cycle, poor specificity, and inability to accurately quantify. While HPLC offers high separation efficiency and quantitative accuracy, its complex sample pretreatment, high detection cost, and slow analysis speed have limited its widespread application. ELISA is simple to operate, moderately costly, and highly sensitive, but it is susceptible to human error and interfering substances, and cross-reactions exist, requiring further improvement in accuracy, although optimization has slightly improved it. MS / MS combines high sensitivity and high specificity, enabling simultaneous detection of multiple components, but the expensive instruments, stringent operational requirements, and high level of expertise required of the personnel hinder its widespread adoption. Sensor detection is convenient to operate and has a fast response speed, enabling real-time monitoring, but its stability is poor, its detection range is limited, it is easily affected by the water matrix, and its detection accuracy is insufficient for monitoring complex water bodies. Long-term practice and research have revealed that these methods are no longer fully adequate for the needs of rapid, accurate, and efficient monitoring of microcystin levels in aquatic bodies in the new era. Therefore, developing a large-scale, convenient, efficient, and accurate method for detecting microcystin levels in aquatic bodies has become an urgent need for scientific researchers.

[0004] In the prior art, Chinese patent document CN117114399A discloses "A method for monitoring the risk of microcystin based on satellite phycocyanin concentration." This method uses the reflectance of nine bands and five spectral indices from an OLCI satellite sensor as input, establishes a remote sensing estimation model for phycocyanin concentration based on an enhanced random forest regression model, and then indirectly retrieves microcystin concentration using phycocyanin as an intermediate variable, ultimately achieving risk level classification. However, this technical solution employs a dual-model indirect strategy, using phycocyanin as an intermediate variable. The model chain is long, resulting in significant error accumulation. Furthermore, the relationship between microcystin and the intermediate variable phycocyanin concentration is unstable in both time and space, reducing the accuracy of the dual-model indirect estimation strategy. This means that the method can only achieve qualitative risk level assessment and cannot accurately retrieve concentration.

[0005] In summary, existing methods for detecting microcystin content in water bodies require high precision instruments, expensive equipment and materials, complex water sample pretreatment processes, and low efficiency. Meanwhile, remote sensing indirect inversion methods suffer from poor accuracy and difficulty in accurately quantifying microcystin. Summary of the Invention

[0006] This invention solves the technical problems of existing technologies, such as high precision requirements for detection instruments, expensive equipment and materials, complex water sample pretreatment process, and low efficiency, as well as the poor accuracy of remote sensing indirect inversion method, which makes it difficult to accurately quantify microcystin.

[0007] The present invention provides a method for estimating the concentration of algal toxins in water bodies based on remote sensing images, comprising the following steps: Step 1: Acquire remote sensing images of the water body and preprocess them to obtain remote sensing reflectance data; Step 2: Calculate the spectral index based on remote sensing reflectance data; Step 3: Construct a remote sensing estimation model for the concentration of microcystin in water. Use the spectral index as the input to the remote sensing estimation model for the concentration of microcystin in water to obtain the concentration of microcystin in water.

[0008] Furthermore, in one embodiment of the present invention, the preprocessing in step 1 specifically includes: Atmospheric correction is performed on the remote sensing image of the water body. The atmospherically corrected image is then separated into land and water, land pixels are masked, and outlier removal is performed on the remote sensing reflectance of the water body pixels to obtain remote sensing reflectance data.

[0009] Furthermore, in one embodiment of the present invention, the spectral indices in step 2 include Rrs_709 / Rrs_620, Rrs_709 / Rrs_490, Rrs_665 / Rrs_620 and (Rrs_779-Rrs_490) / (Rrs_779+Rrs_490).

[0010] Furthermore, in one embodiment of the present invention, the remote sensing estimation model for the concentration of microcystin in water in step 3 is a random forest model.

[0011] Furthermore, in one embodiment of the present invention, the remote sensing estimation model for the concentration of microcystin in water in step 3 is a multiple linear regression model.

[0012] Furthermore, in one embodiment of the present invention, the parameters of the random forest model are set as follows: the number of parameter estimations is 22, the maximum depth is 3, the minimum number of split samples is 11, the minimum number of leaf trees is 5, and the maximum feature is 3.

[0013] Furthermore, in one embodiment of the present invention, the multiple linear regression model is: ; in, This refers to the concentration of algal toxins. The surface reflectance of each band of the OLCI remote sensing image pixel is represented.

[0014] The present invention discloses a system for estimating the concentration of algal toxins in water bodies based on remote sensing images. The system, used to implement the above method, includes the following modules: The preprocessing module acquires remote sensing images of the water body and performs preprocessing to obtain remote sensing reflectance data; The index calculation module calculates spectral indices based on remotely sensed reflectance data. The estimation module constructs a remote sensing estimation model for the concentration of microcystin in water bodies, using spectral indices as input to the model to obtain the concentration of microcystin in water bodies.

[0015] This invention solves the technical problems of existing technologies, such as high precision requirements for detection instruments, expensive equipment and materials, complex water sample pretreatment processes, and low efficiency, while remote sensing indirect inversion methods suffer from poor accuracy and difficulty in accurately quantifying microcystin. Specific beneficial effects of this invention include: This invention proposes a method for estimating the concentration of microcystin in water bodies based on remote sensing images. It can accurately predict the concentration of microcystin using only four specific spectral indices as input. It features wide coverage, convenient operation, and applicability to various types of water bodies. It achieves quantitative inversion of microcystin concentration with extremely simple input, balancing accuracy and efficiency, and is easy to deploy efficiently in lightweight remote sensing ground application systems. Attached Figure Description

[0016] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart of the method for estimating the concentration of algal toxins in water as described in Implementation Method 1; Figure 2 This is a graph showing the estimation accuracy and validation results of the random forest model described in Implementation Method 4. Figure 2 (a) Estimating the training accuracy for the random forest model. Figure 2 (b) Estimating the validation accuracy of the random forest model; Figure 3 This is a graph showing the estimation accuracy and validation results of the multiple linear regression model described in Implementation Method 4. Figure 3 (a) Estimating the training accuracy for a multiple linear regression model. Figure 3 (b) Estimating the validation accuracy of the multiple linear regression model.

[0017] Figure 4This is a comparison chart of the inversion accuracy between the method of the present invention described in Embodiment 4 and the traditional single-band model, wherein... Figure 4 (a) is the accuracy of the band ratio (Rrs_709 / Rrs_620) model estimation. Figure 4 (b) To verify the accuracy of this ratio model, Figure 4 (c) represents the accuracy of the band ratio (Rrs_665 / Rrs_620) model estimation. Figure 4 (d) is the accuracy of the ratio model verification. Figure 4 (e) represents the accuracy of the band ratio (Rrs_709 / Rrs_490) model estimation. Figure 4 (f) is the accuracy verification of this ratio model. Figure 4 (g) represents the accuracy of the normalized exponential model estimation. Figure 4 (h) represents the verification accuracy of the normalized model; Detailed Implementation Various embodiments of the present invention will now be clearly and completely described with reference to the accompanying drawings. The embodiments described with reference to the drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0018] Implementation Method 1: Existing remote sensing methods rely on phycocyanin as an intermediate bridge to indirectly estimate microcystin. This two-stage approach inevitably introduces error accumulation. In addition, the spatiotemporal heterogeneity of the relationship between phycocyanin and microcystin limits the accuracy of quantitative inversion.

[0019] To address the aforementioned technical problems, this embodiment proposes a method for estimating the concentration of microcystin in water bodies based on remote sensing imagery, such as... Figure 1 As shown, this method, using only four specific spectral indices as input, can accurately predict microcystin concentrations. It features wide coverage, ease of operation, and applicability to diverse water body types, providing a new technical approach for large-scale, efficient monitoring of microcystins. Specifically, it includes the following: Step 1: Acquire remote sensing images of the water body and preprocess them to obtain remote sensing reflectance data; (1) Image download: Sentinel 3 OLCI image data (Sentinel 3 satellite ocean and land color imager image data) were selected based on the study area, time and resolution, and images with less cloud cover and higher quality were selected from the official platform.

[0020] (2) Atmospheric correction: The ACOLITE atmospheric correction method (water area atmospheric correction tool) is used to eliminate atmospheric scattering, absorption and other interferences, and convert the apparent reflectance into the true surface reflectance.

[0021] (3) Reflectance extraction: After cropping and masking to remove noise such as land and cloud shadows, the reflectance of the water body band is extracted and statistical and outlier processing is performed to provide reliable data for the subsequent construction of the microcystin concentration remote sensing model.

[0022] Step 2: Calculate the spectral index based on remote sensing reflectance data; Existing remote sensing indirect inversion methods involve more than ten input variables, resulting in high redundancy of band information, large model complexity, and a lack of mechanistic drivers for variable selection. The large number of redundant variables not only increases the computational burden but also forces the scheme to rely on complex machine learning models, making it difficult to deploy in lightweight remote sensing systems.

[0023] To address the aforementioned issues, this implementation method uses only four spectral indices to achieve quantitative inversion of microcystin concentration with extremely simple input, balancing accuracy and efficiency.

[0024] Existing technologies for screening characteristic bands for microcystin concentration inversion mostly use correlation analysis, which involves ranking all bands by correlation coefficients in single bands or various combinations of addition, subtraction, multiplication, and division. This approach does not consider factors such as the microcystin production process, biochemical mechanisms, water component correlations, and optical properties. It lacks analysis at the theoretical level of bio-optical remote sensing, resulting in unique characteristics for each region, poor model universality, and an inability to determine universal characteristic bands that can effectively serve the inversion of microcystin concentrations in various types of water bodies, making it difficult to promote and apply.

[0025] To overcome the above difficulties, this embodiment starts with the optical information required for the inversion of microcystin concentration. Based on the differentiated optical response characteristics of various characteristic pigments of toxin-producing cyanobacteria and background components of water in the visible-near infrared band, and with the help of correlation analysis to verify, the characteristic bands are determined to be Rrs_490, Rrs_620, Rrs_665, Rrs_709 and Rrs_779.

[0026] To highlight the remote sensing signals of phycocyanin and chlorophyll a, and simultaneously reduce the interference from topography, illumination, observation angle, and other water components, this embodiment uses four band combinations "Rrs_709 / Rrs620", "Rrs_665 / Rrs620", "Rrs_709 / Rrs_490", and "(Rrs_779-Rrs_490) / (Rrs_779+Rrs_490)" as sensitive band combinations for algal toxin concentration, and as input variables for the remote sensing estimation model of water microcystin concentration. The spectral indices Rrs_709 / Rrs_620 utilize the ratio of the 709 nm reflection peak (generated by high concentrations of algae) to the 620 nm absorption valley (caused by specific absorption of phycocyanin) to effectively amplify the phycocyanin signal; Rrs_709 / Rrs_490 is used to eliminate the influence of some water background (such as colored dissolved organic matter), thereby highlighting the concentration information of chlorophyll a; Rrs_665 / Rrs_620 is used to effectively offset the common background interference from other components, making the phycocyanin signal purer; (Rrs_779-Rrs_490) / (Rrs_779+Rrs_490) is used to maximize the highlighting of high-density algae water bodies and effectively reduce the influence of external environmental factors such as light and observation angle.

[0027] Step 3: Construct a remote sensing estimation model for the concentration of microcystin in water. Use the spectral index as the input to the remote sensing estimation model for the concentration of microcystin in water to obtain the concentration of microcystin in water.

[0028] This embodiment utilizes an ELISA kit (enzyme-linked immunosorbent assay kit) to test the concentration of microcystin in water samples (lake samples only), primarily employing biological methods for testing. The experimental steps are as follows: (1) Shake the water sample evenly, take an appropriate amount of water sample containing algal particles, and repeatedly freeze and thaw to destroy its cell wall, so that the intracellular microcystin toxin is fully released into the water sample.

[0029] (2) The above water sample was filtered using a GF / F filter membrane (glass fiber filter membrane) with a pore size of 0.45 μm, retaining a portion of the filtrate. The concentration standard curve of microcystin (R) was then used to analyze the microcystin concentration. 2 (Up to 0.999), the microcystin content in the filtrate was measured using an enzyme-linked immunosorbent assay (ELISA) reader, and the microcystin concentration of each water sample was determined sequentially according to this method.

[0030] Spectral indices of the pixel locations in remote sensing images of test water samples were obtained. Combined with the measured microcystin concentrations (using an ELISA kit), a sample dataset was constructed and proportionally divided into training and validation sets. A remote sensing estimation model for water microcystin concentration was trained using the four spectral indices as input and the measured microcystin concentration as output. After training, the model accuracy was evaluated using the validation set, yielding the final remote sensing estimation model for water microcystin concentration.

[0031] Implementation Method 2: The difference between this implementation method and Implementation Method 1 is that the remote sensing estimation model for the concentration of microcystin in water in step 3 is a random forest model.

[0032] Using the spectral index described in Implementation Method 1 as the input variable for the random forest algorithm, the model parameters were set to 22 estimation iterations, a maximum depth of 3, a minimum number of split samples of 11, a minimum number of leaf trees of 5, and a maximum feature of 3. A remote sensing estimation model for microcystin concentration in water was constructed. The results show that the estimation accuracy R0 is... 2 =0.8, RMSE=0.32µg·L -1 , where R 2 The coefficient of determination is RMSE, which is the root mean square error.

[0033] Implementation Method 3: The difference between this implementation method and Implementation Method 1 is that the remote sensing estimation model for the concentration of microcystin in water in step 3 is a multiple linear regression model.

[0034] Using the spectral index described in Implementation Method 1 as the independent variable of a multiple linear regression model and the measured microcystin concentration as the dependent variable, a multiple linear regression equation y = Ax1 + Bx2 + Cx3 + Dx4 + b is established. The coefficients of this equation are obtained by using the least squares method on 54 pairs of sample data: A = 0.451, B = -0.248, C = 9.82, D ​​= 1.52, and b = -6.98. Here, y represents the microcystin concentration, and x1 to x4 represent the combinations of the four sensitive spectral bands mentioned above. Therefore, the method for estimating the concentration of microcystin in water bodies in this implementation method is as follows: ; in, This refers to the concentration of algal toxins, in µg·L. -1 , The surface reflectance of each band in the OLCI remote sensing image is expressed in Sr. -1 .

[0035] In summary, Implementation Methods 2 and 3 utilize satellite imagery data to construct a remote sensing estimation model for microcystin concentration in water bodies, thereby estimating the concentration of microcystins. This means that only satellite imagery data of the target water body needs to be acquired, input into the model, and the model parameters simultaneously set; the microcystin concentration of the target water body can then be estimated efficiently. Furthermore, the random forest model achieves an estimation accuracy of up to R0. 2 =0.8, RMSE=0.32µg·L -1 The accuracy of the multiple linear regression model is as high as R0. 2 =0.63, RMSE=0.44µg·L -1 It has high reliability, balances accuracy and efficiency, and can be used in the field of large-scale water quality monitoring and evaluation, providing decision support for water environment protection departments.

[0036] Implementation Method Four: This implementation method is a specific embodiment proposed for the methods described in Implementation Methods One to Three.

[0037] Sampling was conducted at Gehu Lake, Taihu Lake, and Dianchi Lake, with 20-40 sampling points set up at each lake, totaling 79 sampling points. Surface water samples (0.5-1 m) were collected using a water sampler, with 2 L of surface water (0.5-1 m) collected at each sampling point. The samples were transported to the laboratory in a vehicle refrigerator and stored under light-protected conditions. Microcystin concentrations were determined within 24 hours of arrival at the laboratory. The assay procedure was as follows: Standard / test solution and enzyme-labeled antibody were added to microplates and incubated at 37°C in the dark for 30 min. The plates were washed 5 times, incubated with chromogenic solution for 15 min, and the reaction was terminated with stop solution. OD values ​​were measured at 450 nm using a microplate reader, a standard curve was plotted, and the MC concentration was calculated.

[0038] This method requires instruments costing approximately 18,000-28,000 yuan, and operation by experienced laboratory technicians. Due to human error, labor costs are 300-500 yuan per day, totaling 291-438 yuan per sample. The total cost per sample is 5200-7100 yuan. Each sample takes approximately 15-20 minutes, and the sampling time is 2 days, for a total of approximately 3 days.

[0039] Image download, atmospheric correction, inputting the latitude and longitude of the collected water sample into the image, and extracting the reflectance of the pixel at that point: (1) Remote sensing image processing is the foundation for subsequent microcystin concentration estimation. The process includes key steps such as image download, atmospheric correction, and reflectance extraction. First, in the image download stage, Sentinel 3 OLCI images are downloaded from the European Space Agency website (https: / / ladsweb.modaps.eosdis.nasa.gov / search / ) based on the size, spatial location, and monitoring date of the target water body. During the download process, images from 7 days before and after the monitoring date can be selected based on factors such as cloud cover and imaging weather to ensure data quality.

[0040] (2) The downloaded images were subjected to ACOLITE atmospheric correction. The key steps are as follows: the original digital number (DN) values ​​were converted into top atmospheric layer (TOA) reflectance or radiance; the Dark Spectrum Fitting (DSF) algorithm was used to estimate the aerosol optical thickness (AOT); Rayleigh scattering correction was used to eliminate the influence of molecular scattering on short-wave bands (such as blue light); diffuse transmittance was calculated to compensate for the interference of atmospheric path radiation on the surface reflection signal; for the specular reflection of the water surface, flare correction was enabled under coastal or strong light conditions to finally obtain the true spectral characteristics of the water surface.

[0041] (3) Reflectance extraction: After obtaining the surface reflectance image through atmospheric correction, the reflectance information of the water body in the study area was accurately extracted in each visible and near-infrared band using SeaDAS software (Marine Color Remote Sensing Data Processing System) for band clipping, masking, and other methods. Outliers were removed.

[0042] The satellite imagery data, batch processing code for the ACOLITE atmospheric correction algorithm, and reflectance extraction software used in this operation are all available free of charge online. Operational tutorials for each step can also be found in the website's free resources. This process is zero-cost, takes approximately 2 hours, and saves 99% of the cost.

[0043] In this embodiment, four band combinations, “Rrs_709 / Rrs620”, “Rrs_665 / Rrs620”, “Rrs_709 / Rrs_490”, and “(Rrs_779-Rrs_490) / (Rrs_779+Rrs_490)”, are used as input variables for the remote sensing estimation model of microcystin concentration in water.

[0044] To ensure that the above band combinations also have a strong response to microcystin concentrations, this embodiment uses correlation analysis to calculate the correlation coefficients between the four band combinations and the measured microcystin concentrations to verify their sensitivity. The correlation coefficients (r) between each band combination and the microcystin concentration were found to be greater than 0.65, indicating that the above band combinations meet the requirements for estimating microcystin concentrations in water bodies.

[0045] Example 1: Using the above four sensitive band combinations as input variables for the random forest model has advantages over single bands or other band combinations, including weak correlation removal, high estimation accuracy, and strong robustness. Furthermore, three of the four sensitive band combinations are ratio combinations, which automatically eliminate interference from water noise, atmosphere, and light. These four combinations complement each other as optimal features, making the random forest estimation model for microcystin faster, more accurate, and more stable.

[0046] The four selected sensitive band combinations are input into the random forest algorithm. The model accuracy is validated using a 5-fold spectral density test, and the training and validation accuracy of the model are plotted in Origin (Origin scientific plotting and data analysis software). Figure 2 (a) shows the training accuracy, as shown in Figure 1. Figure 2 (b) shows the validation accuracy. The training accuracy of this model is R0. 2 =0.88, RMSE=0.25µg·L -1 The verification accuracy is R. 2 =0.9, RMSE=0.20µg·L -1 This indicates that the random forest remote sensing model can effectively predict the concentration of microcystin in water using remote sensing data, thus achieving accurate estimation of the microcystin content in water.

[0047] This estimation method can directly estimate the microcystin content in water bodies by extracting spectral information from satellite imagery. The process incorporates remote sensing principles and correlation analysis to screen sensitive features, amplify the signal, remove redundancy, eliminate interference, and generate a superposition effect, thus improving estimation accuracy. Compared to traditional single-band combined estimation models, the accuracy is approximately 3-4 times higher. Figure 4 As shown.

[0048] This embodiment improves the accuracy of microcystin concentration estimation in water by 66%, enabling large-scale water quality monitoring. Compared with traditional monitoring methods, it saves 99% in costs and increases work efficiency by 35%, which is beneficial to supporting water environment protection.

[0049] Example 2: Using the above four sensitive band combinations as independent variables in a multiple linear regression model has advantages over single bands or other band combinations, including the removal of weak correlations, higher estimation accuracy, and stronger robustness. Furthermore, three of the four sensitive band combinations are ratio combinations, which automatically eliminate interference from water noise, atmosphere, and light. These four combinations complement each other as optimal features, making the multiple linear regression estimation model for microcystin faster, more accurate, and more stable.

[0050] Using the multiple linear regression fitting tool in Origin software, the measured microcystin concentrations at 54 sampling points were matched one-to-one with the characteristic values ​​of each sampling point in four band combinations. These values ​​were then input into the multiple linear regression function y = Ax1 + Bx2 + Cx3 + Dx4 + b. The optimal coefficients were calculated using the least squares method: A = 0.451, B = -0.248, C = 9.82, D ​​= 1.52, and b = -6.98. The goodness of fit of the model was obtained as R² = 0.63, and the relative root mean square error (RMSE) was 0.49 µg·L⁻¹. -1 ,like Figure 3 As shown in (a), based on this model, the concentration of microcystin in water can be calculated using the reflectance extracted from satellite images.

[0051] To verify the accuracy of the estimation results, this embodiment simultaneously measured the microcystin concentration in 25 water samples in the field. These measured microcystin values ​​were then fitted and analyzed with the microcystin concentrations calculated according to the method of this embodiment to verify the model's accuracy. Figure 3 As shown in (b), the results indicate that the relative root mean square error (RMSE) is only 0.44 µg·L⁻¹. -1 The ratio of the measured microcystin concentration to the calculated value in this embodiment is 0.90, demonstrating that the concentration of microcystins in water can be accurately estimated using a combination of reflectance in specific bands of remote sensing images. The microcystin concentration calculated using the method in this embodiment has high estimation accuracy and reliability. The multiple linear regression model between the combination of reflectance in specific bands of remote sensing images and the microcystin concentration was first established in this embodiment.

[0052] This method estimates the microcystin content in water bodies by extracting spectral information from satellite imagery. It incorporates remote sensing principles and correlation analysis to screen sensitive features, amplify signals, remove redundancy, eliminate interference, and generate a superposition effect, thus improving estimation accuracy. Compared to traditional single-band model estimation methods, the accuracy is 2-3 times higher. Figure 4 ).

[0053] This embodiment improves the accuracy of microcystin concentration estimation in water bodies by 33%, enabling large-scale monitoring of microcystin concentration in water bodies. Compared with traditional monitoring methods, it saves 99% of costs and increases work efficiency by 35%, which is beneficial to supporting the monitoring and protection of aquatic ecological health.

[0054] This implementation method constructs a remote sensing estimation model for microcystin concentration in water bodies using satellite imagery data, establishing a novel method for estimating microcystin content in water bodies. This method requires only acquiring satellite images of the target water body to estimate the microcystin concentration using this remote sensing model. It boasts high accuracy, high reliability, low cost, and high efficiency, and can be applied to large-scale water quality monitoring and evaluation, providing decision support for water environment protection departments.

[0055] The above provides a detailed description of the method and system for estimating the concentration of algal toxins in water based on remote sensing images proposed in this invention. Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A method for estimating the concentration of algal toxins in water bodies based on remote sensing imagery, characterized in that, Includes the following steps: Step 1: Acquire remote sensing images of the water body and preprocess them to obtain remote sensing reflectance data; Step 2: Calculate the spectral index based on remote sensing reflectance data; Step 3: Construct a remote sensing estimation model for the concentration of microcystin in water. Use the spectral index as the input to the remote sensing estimation model for the concentration of microcystin in water to obtain the concentration of microcystin in water.

2. The method for estimating the concentration of algal toxins in water bodies based on remote sensing imagery according to claim 1, characterized in that, The preprocessing in step 1 specifically includes: Atmospheric correction is performed on the remote sensing image of the water body. The atmospherically corrected image is then separated into land and water, land pixels are masked, and outlier removal is performed on the remote sensing reflectance of the water body pixels to obtain remote sensing reflectance data.

3. The method for estimating the concentration of algal toxins in water bodies based on remote sensing imagery according to claim 1, characterized in that, In step 2, the spectral indices include Rrs_709 / Rrs_620, Rrs_709 / Rrs_490, Rrs_665 / Rrs_620 and (Rrs_779-Rrs_490) / (Rrs_779+Rrs_490).

4. The method for estimating the concentration of algal toxins in water bodies based on remote sensing imagery according to claim 1, characterized in that, In step 3, the remote sensing estimation model for the concentration of microcystin in the water is a random forest model.

5. The method for estimating the concentration of algal toxins in water bodies based on remote sensing imagery according to claim 1, characterized in that, In step 3, the remote sensing estimation model for the concentration of microcystin in water is a multiple linear regression model.

6. The method for estimating the concentration of algal toxins in water bodies based on remote sensing imagery according to claim 4, characterized in that, The parameters of the random forest model are set as follows: parameter estimation times are 22, maximum depth is 3, minimum number of split samples is 11, minimum number of leaf trees is 5, and maximum feature is 3.

7. The method for estimating the concentration of algal toxins in water bodies based on remote sensing imagery according to claim 5, characterized in that, The multiple linear regression model is as follows: ； in, This refers to the concentration of algal toxins. The surface reflectance of each band of the OLCI remote sensing image pixel is represented.

8. A system for estimating the concentration of algal toxins in water bodies based on remote sensing imagery, the system being used to implement the method of claim 1, characterized in that, Includes the following modules: The preprocessing module acquires remote sensing images of the water body and performs preprocessing to obtain remote sensing reflectance data. The index calculation module calculates spectral indices based on remotely sensed reflectance data. The estimation module constructs a remote sensing estimation model for the concentration of microcystin in water. The spectral index is used as the input to the remote sensing estimation model for the concentration of microcystin in water to obtain the concentration of microcystin in water.

Images