An Unmanned Aerial Vehicle Hyperspectral Remote Sensing Detection Method for Water Pollution Sources

Abstract

The present invention discloses a method for detecting water pollution sources by using an unmanned aerial vehicle (UAV) hyperspectral remote sensing, belonging to the technical field of water environment monitoring. The method includes obtaining hyperspectral image data of water bodies through UAV flight and synchronously collecting ground water quality data, performing general preprocessing on the obtained image data, extracting spectral information of corresponding ground-measured points according to the results of the image data preprocessing, establishing a quantitative mathematical model of water quality and hyperspectral data and validating the model, using the model to realize remote sensing inversion of water quality and visualization of concentration spatial distribution data, extracting concentration anomaly points and performing visual analysis on the fluid characteristics of concentration diffusion distribution, so as to confirm the types and locations of pollution discharges. When using the UAV hyperspectral water pollutant detection technology to quickly and accurately conduct a general survey of water eutrophication and pollution degree and accurately locate the positions of pollution outlets, it can realize accurate monitoring of the water quality environment and carry out targeted prevention and control.

Description

Technical Field

[0001] The present invention relates to the technical field of water environment monitoring, and in particular to a method for detecting water pollution sources using unmanned aerial vehicle (UAV) hyperspectral remote sensing. Background Art

[0002] With the rapid development of human society, pollution discharge caused by human activities has become a very common phenomenon, with the consequence of serious eutrophication of water bodies, and even black and smelly water bodies, which has caused drinking water safety and water resource shortages. For sewage discharge, all countries have formulated strict monitoring measures and severe law enforcement measures as well as a large number of ecological governance measures, but the phenomenon of illegal discharge of pollution is still rampant, and the means and forms are becoming more and more diverse, especially the illegal discharge of underground pipes, leakage and so on. Traditional methods such as manual inspections and point source sampling detection can generally only detect the pollution of open discharge of sewage, but these hidden and "hide-and-seek" illegal discharge methods cannot be discovered, and must rely on some new technical means to achieve.

[0003] Among them, the UAV hyperspectral imaging water pollutant screening technology is a new water quality measurement technology combined with hyperspectral remote sensing. This technology uses a drone equipped with a hyperspectral imager to perform remote sensing scanning imaging to obtain images and spectral information of ground objects (including pollutants in water bodies), and then uses a model algorithm to analyze the spectral characteristics of different objects to obtain the qualitative or quantitative relationship between the measurement parameters and the spectrum, thereby analyzing the data of the measured parameters. By analyzing this data, the type of pollutants in the water body and whether pollution is discharged can be determined.

[0004] Compared with traditional satellite remote sensing technology, UAV hyperspectral imaging remote sensing technology has the large-scale surface measurement characteristics of satellite remote sensing, and also has finer spatial resolution and temporal flexibility than satellite remote sensing. Its hyperspectral has more spectral bands and finer spectral resolution than satellite multi-spectral, and can better characterize the characteristic spectra of water bodies and other land objects.

[0005] Therefore, the UAV hyperspectral water pollutant detection technology has obvious advantages and advancements in quickly and accurately surveying the eutrophication and pollution levels of water bodies and finding pollution outlets. The establishment of this technical method has important significance and theoretical value for water quality environment monitoring and prevention. Summary of the invention

[0006] The purpose of the present invention is to provide a method for detecting water pollution sources by unmanned aerial vehicle hyperspectral remote sensing to solve the problems raised in the above-mentioned background technology.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions: a method for detecting water pollution sources by using a drone with high-spectral remote sensing, the method comprising the following steps:

[0008] S1. Use drone aerial flight to obtain hyperspectral image data of water bodies and simultaneously collect ground water quality data;

[0009] S2. Preprocess the obtained image data;

[0010] S3. Extract spectral information of corresponding ground measured points according to the results of image data preprocessing;

[0011] S4. Establish a quantitative mathematical model of water quality and hyperspectral data and verify the model;

[0012] S5. Use the model to realize remote sensing inversion of water quality and visualization of concentration spatial distribution data;

[0013] S6. Extract concentration anomaly points and conduct visual analysis on the fluid characteristics of concentration diffusion distribution;

[0014] S7. Confirm the sewage discharge type and location.

[0015] According to the above technical solution, in S1, the use of drone aerial flight to obtain hyperspectral image data of water bodies means using drone hyperspectral imaging remote sensing technology to obtain hyperspectral remote sensing image data of the target water area; before each flight, lay out a diffuse reflection gray cloth with a reflectance of 20%-40% or a diffuse reflection white cloth with a reflectance of more than 97% on the flight route as a reference target; the size of the reference target should at least completely occupy an image element area of at least 10×10 in the spectral image to ensure sufficient effective data, and each flight route must pass over the reference target.

[0016] The collection of ground water quality data refers to selecting representative sample points, obtaining water samples through manual sampling, and bringing them back to the laboratory to detect the concentration of water quality indicators according to the analysis methods of the Surface Water Environment Quality Standard GB 3838-2002. The water quality indicators refer to total nitrogen, total phosphorus, chlorophyll a, permanganate index, ammonia nitrogen, chemical oxygen demand, etc.

[0017] According to the above technical solution, in S2, the preprocessing includes geometric correction, radiometric calibration, image splicing, and water area extraction; after preprocessing, remote sensing image data with standard hyperspectral remote sensing reflectance can be obtained;

[0018] The geometric correction refers to obtaining a high-resolution satellite geographic image with a geographic coordinate system in the area where the target water area is located and that can be publicly obtained, using it as a reference image for geometric correction, and using the geometric correction tool of remote sensing software to perform "image-image" geometric correction by the method of "homologous point matching" to ensure the correctness of the acquired image;

[0019] The radiation calibration refers to finding the data area of the reference target on the UAV remote sensing image, extracting the reflectance data of more than 3 target reference plates, and then taking the average value;

[0020] The calculation formula for the remote sensing reflectance data of the target reference plate is:

[0021]

[0022] Then, the spectral remote sensing DN value data of each pixel is calibrated by the reflectance-based method to obtain the UAV spectral remote sensing reflectance data. The calculation formula of the reflectance-based method is:

[0023]

[0024] represents the calculated UAV remote sensing reflectance data (sr -1 ), R ref (λ) represents the reflectance value of the actually measured target reference plate, represents the calculated remote sensing reflectance value of the target reference plate (sr -1 ), DN UAV (λ) represents the spectral signal value measured by the UAV, and DN ref (λ) represents the spectral signal value of the target reference plate detected by the UAV;

[0025] The image mosaicking and cropping refers to automatically mosaicking using the mosaicking tool of remote sensing software and manually polygon cropping using the cropping tool of remote sensing software; The reason for image acquisition is that UAV remote sensing images are generally stored in tiles, and the subsequent image processing should be mosaicked to form a complete water body image; In addition, through cropping, unnecessary areas can be removed, which simplifies the analysis process to a certain extent;

[0026] The water area extraction refers to extraction by setting thresholds according to the water body characteristic bands. The water body threshold calculation formula is:

[0027] NDVI = (B1 - B2) / (B1 + B2);

[0028] NDWI = (B3 - B4) / (B3 + B4);

[0029] Among them, both B1 and B3 are the reflectances of the red light band, B2 and B4 are the reflectances of the near-infrared light band, NDVI is the normalized difference vegetation index, and NDWI is the normalized difference water index; The NDWI threshold of the water area is generally greater than 0.1 - 0.3, and the specific threshold needs to be determined by checking the accuracy of the land-water boundary extraction; For the extracted land-water boundary line, the error is generally required to be no more than 3 pixels;

[0030] Then, the calculated threshold is extracted to generate a mask, and the mask is converted into a vector file of the water-land boundary through the ROI tool. Finally, the ROI tool can be used to crop and extract the water body based on the vector file, and the hyperspectral image data of the water area is obtained.

[0031] According to the above technical solution, in S3, the step of extracting the spectral information of the corresponding ground measured point is as follows:

[0032] L1. Input the longitude and latitude of the ground water quality measurement point into a text in TXT format to generate a site longitude and latitude file;

[0033] L2. Open the target image data and gradually import the corresponding file using the ROI from ASCII file tool;

[0034] L3. Then click the region of interest tool and select export to CSV to output the result in CSV format.

[0035] According to the above technical solution, in S4, the establishment of a quantitative mathematical model for water quality and hyperspectral data refers to using the water quality data collected in the field and the hyperspectral data of the unmanned aerial vehicle for mathematical modeling to establish a quantitative relationship between water quality and hyperspectral;

[0036] The method for establishing the quantitative mathematical model is as follows:

[0037] Establish empirical and semi-empirical models, including linear correlation models, polynomial models, and multiple regression models;

[0038] The model establishment formulas are as follows:

[0039] Linear correlation model: N = a×M + b;

[0040] Polynomial correlation model: N = a×M 2 + b×M + C;

[0041] Multiple regression model: N = a 0 + a 1 B 1 + a 2 B 2 + …… + a i B i + d;

[0042] M is or a and b are regression coefficients; N is the concentration value of the water quality index, a0, a1, …… ai are coefficients, d is the error term, and B1, B2, …… Bi are the reflectance values of the hyperspectral bands.

[0043] According to the above technical solution, in S4, the quantitative mathematical model building method further includes an artificial neural network model, and the specific steps are as follows:

[0044] X1. Randomly select 60% of the data, use the reflectance of each band as the input layer to perform network training with the measured water quality values, and construct a stable network;

[0045] X2. Then use the remaining 40% of the data to verify the accuracy of the model;

[0046] X3. By adjusting the weights and optimizing the model structure until the accuracy meets the requirements, the training ends and a stable model is formed;

[0047] The establishment of multiple quantitative mathematical models makes the results of data processing more accurate.

[0048] According to the above technical solution, in S4, the calibration model is to calibrate the water quality remote sensing inversion model;

[0049] The calibration of the water quality remote sensing inversion model is used to test the practicability of the model, making the data analysis of the model more accurate;

[0050] The calibration of the water quality remote sensing inversion model is evaluated using the following formula:

[0051]

[0052]

[0053] Where RMSE in the formula represents the root mean square error, MAPE represents the mean percentage error, and respectively represent the measured value and the predicted value of the water quality index at point i, and n represents the number of samples.

[0054] According to the above technical solution, in S5, the steps of water quality remote sensing inversion and data visualization are as follows:

[0055] Z1. Use the data model established in S4 and import the model into the calculation tool;

[0056] Z2. Calculate step by step according to the window prompts of the tool to obtain the water quality hyperspectral remote sensing inversion data;

[0057] Z3. Classify the water quality hyperspectral remote sensing inversion data according to the data step level, assign different colors to each level, and save the output;

[0058] The data is generated into a visualization map that uses different colors to indicate the water quality concentration range and spatial distribution characteristics; this map can be overlaid with a geographic base map to generate a remote sensing result map containing additional external geographic environment characteristics of the water area, making the data more intuitive.

[0059] According to the above technical solution, in S6, the extraction of the concentration anomaly point is obtained by analyzing the visualization data of the concentration distribution of each indicator, and the morphological characteristics of the gradient distribution of the concentration are depicted with the anomaly point as the center point. This characteristic is the concentration diffusion characteristic of the pollutant. The diffusion of the pollutant concentration is manifested as a gradient decrease from the sewage outlet to the surrounding area. The fluid characteristics are generally point-shaped, block-shaped, diffuse, or band-shaped or line-shaped diffusion along the direction of the water flow; this allows a clearer understanding of the concentration diffusion of pollutants.

[0060] According to the above technical solution, in S7, the confirmation of the sewage discharge location refers to the point with the highest concentration, that is, the location of the sewage outlet, based on the pollutant diffusion characteristics. The longitude and latitude of the sewage outlet are recorded according to the geographical location information provided by the image data. At the same time, it can clearly indicate the water quality indicators of the polluted discharge, that is, determine the type of pollution discharged, so that the degree of water pollution can be quickly and accurately surveyed and the pollution outlet can be found.

[0061] Through the above technical solution, a rapid and accurate survey of the eutrophication and pollution level of water bodies can be carried out, the pollution outlets can be accurately located, and the diffusion of pollutant concentrations can be more intuitively understood, so that water quality environment monitoring can be carried out in a timely manner and targeted prevention and control can be carried out.

[0062] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0063] 1. The present invention utilizes a UAV equipped with a hyperspectral imager for remote sensing scanning imaging, which enables measurement of a large area, and has finer spatial resolution and time flexibility than satellite remote sensing. At the same time, its hyperspectral image has more spectral bands and finer spectral resolution than the satellite's multi-spectral image, and can better characterize the characteristic spectra of water bodies and other landforms.

[0064] 2. The present invention utilizes the UAV hyperspectral water pollutant screening technology to quickly and accurately survey the eutrophication and pollution level of water bodies and find pollution outlets, which has obvious advantages and advancements, enabling timely monitoring of water quality environment and targeted prevention and control.

[0065] 3. The present invention utilizes a quantitative mathematical model to parse out the data of the measured parameters, and by analyzing the data, the type of water pollutants and whether sewage is discharged can be determined, thereby making data analysis more accurate and monitoring of sewage more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0067] Figure 1 is the original image of UAV hyperspectral remote sensing;

[0068] Figure 2 is the image raster map generated by calculating the NDWI threshold;

[0069] Figure 3 is the water area remote sensing map after ROI cropping using the vector boundary;

[0070] Figure 4 is the remote sensing result map containing additional geographical environment features outside the water area;

[0071] Figure 5 is the schematic diagram of the typical fluidics form of pollution concentration diffusion;

[0072] Figure 6 is the schematic diagram of UAV hyperspectral remote sensing detection of pollutant emissions;

[0073] Figure 7 is the schematic diagram of the technical process for UAV hyperspectral remote sensing investigation of water pollution sources in the present invention;

[0074] Figure 8 is the schematic diagram of the step process of the method for UAV hyperspectral remote sensing investigation of water pollution sources in the present invention;

[0075] Figure 9 is the working map for the investigation and application along the coast of Qinghai Lake;

[0076] Figure 10 is the working map for the investigation and application in Liangxi River, Wuxi City;

[0077] Figure 11 is the working map for the investigation and application in West Lake, Dali Prefecture, Yunnan;

[0078] Figure 12 is the working map for the investigation and application in Huotong River, Ningde, Fujian;

[0079] Figure 13 is the working map for water sample collection;

[0080] Figure 14 is the working map for water quality detection; Detailed implementation manners

[0081] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0082] Such asFigures 1 to 8 As shown in the figure, the present invention provides a technical solution, a method for detecting water pollution sources by drone hyperspectral remote sensing. The method for detecting water pollution sources by drone hyperspectral remote sensing includes the following steps:

[0083] S1. Use drone flight to obtain hyperspectral image data of the water body and synchronously collect ground water quality data;

[0084] S2. Preprocess the obtained image data;

[0085] S3. Extract spectral information of corresponding ground measured points according to the results of image data preprocessing;

[0086] S4. Establish a quantitative mathematical model of water quality and hyperspectral data and verify the model;

[0087] S5. Use the model to realize remote sensing inversion of water quality and visualization of concentration spatial distribution data;

[0088] S6. Extract concentration anomaly points and conduct visual analysis on the fluid characteristics of concentration diffusion distribution;

[0089] S7. Confirm the pollution type and pollution location.

[0090] In S1, the use of drone flight to obtain hyperspectral image data of the water body refers to using drone hyperspectral imaging remote sensing technology to obtain hyperspectral remote sensing image data of the target water area; before each flight, a diffuse reflection gray cloth with a reflectivity of 20%-40% or a diffuse reflection white cloth with a reflectivity of more than 97% is laid flat on the flight path as a reference target; the size of the reference target should at least completely occupy at least a 10×10 pixel area in the spectral image to ensure sufficient valid data, and each flight path must pass over the reference target; for example, using drone hyperspectral imaging remote sensing technology to obtain hyperspectral remote sensing image data of the target water area, that is, the original drone hyperspectral remote sensing image is as Figure 1 shown.

[0091] The ground water quality data collection refers to selecting representative sample points, obtaining water samples through manual sampling, and bringing them back to the laboratory to detect the concentration of water quality indicators according to the analysis method of the surface water environmental quality standard GB 3838-2002. The water quality indicators refer to total nitrogen, total phosphorus, chlorophyll a, permanganate index, ammonia nitrogen, chemical oxygen demand, etc.

[0092] In S2, the preprocessing includes geometric correction, radiometric calibration, image splicing, and water area extraction; after preprocessing, remote sensing image data with standard hyperspectral remote sensing reflectivity can be obtained;

[0093] The geometric correction refers to obtaining a high-resolution satellite geographic image of the area where the target water area is located, which has a geographic coordinate system and can be publicly obtained, as the reference image for geometric correction, and using the geometric correction tool of remote sensing software to perform "image-image" geometric correction by the method of "homologous point matching" to confirm the correctness of the acquired image. For example, use the Geometric Correction tool in ENVI to perform "image-image" geometric correction;

[0094] The radiometric calibration refers to finding the data area of the reference target on the UAV remote sensing image, extracting the reflectance data of more than 3 target reference plates, and then taking the average value;

[0095] The calculation formula for the remote sensing reflectance data of the target reference plate is:

[0096]

[0097] Then, calibrate the spectral remote sensing DN value data of each pixel by the reflectance-based method to obtain the UAV spectral remote sensing reflectance data. The calculation formula of the reflectance-based method is:

[0098]

[0099] represents the calculated UAV remote sensing reflectance data (sr -1 ), R ref (λ) represents the reflectance value of the actually measured target reference plate, represents the calculated remote sensing reflectance value of the target reference plate (sr -1 ), DN UAV (λ) represents the spectral signal value measured by the UAV, DN ref (λ) represents the spectral signal value of the target reference plate detected by the UAV;

[0100] The image mosaicking and cropping refers to automatically mosaicking using the mosaicking tool of remote sensing software and manually polygon cropping using the cropping tool of remote sensing software; The reason for image acquisition is that UAV remote sensing images are generally stored in tiles, and the subsequent processing of the images should be spliced to form a complete water body image; In addition, through cropping, unnecessary areas can be removed, which simplifies the analysis process to a certain extent; For example, use the Seamless Mosaic tool in ENVI to achieve automatic mosaicking, and use the ROIs tool in ENVI to achieve manual polygon cropping;

[0101] The water area extraction refers to extracting by setting a threshold according to the water body characteristic band. The water body threshold calculation formula is:

[0102] NDVI = (B1 - B2) / (B1 + B2);

[0103] NDWI = (B3 - B4) / (B3 + B4);

[0104] Where B1 and B3 are both reflectance in the red light band, B2 and B4 are reflectance in the near-infrared light band, NDVI is the normalized difference vegetation index, and NDWI is the normalized difference water index; the NDWI threshold for water areas is generally greater than 0.1 - 0.3, and the specific threshold determination requires checking the accuracy of water-land boundary extraction; for the extracted water-land boundary line, the error is generally required not to exceed 3 pixels;

[0105] Then extract the calculated threshold, generate a mask, and then convert the mask into a vector file of the water-land boundary through the ROI tool. Finally, the ROI tool can be used through the vector file to achieve the clipping and extraction of water bodies, obtaining the hyperspectral image data of water areas. For example, the image raster map generated by calculating the NDWI threshold is as Figure 2 shown, and the remote sensing map of water areas after ROI clipping using the vector boundary is as Figure 3 shown.

[0106] In S3, the steps for extracting the spectral information of the corresponding ground-measured points are as follows:

[0107] L1. Input the longitude and latitude of the ground water quality measurement points into a text in TXT format to generate a site longitude and latitude file;

[0108] L2. Open the target image data and gradually import the corresponding files using the ROI from ASCII file tool;

[0109] L3. Then click the region of interest tool to select export to CSV and output the result in CSV format.

[0110] In S4, establishing the quantitative mathematical model between water quality and hyperspectral data means using the water quality data collected in the field and the hyperspectral data of the unmanned aerial vehicle for mathematical modeling to establish the quantitative relationship between water quality and hyperspectral;

[0111] The modeling method for establishing the quantitative mathematical model is as follows:

[0112] Establish empirical and semi-empirical models, including linear correlation models, polynomial models, and multiple regression models;

[0113] The model establishment formulas are as follows:

[0114] Linear correlation model: N = a × M + b;

[0115] Polynomial correlation model: N = a × M 2 + b × M + C;

[0116] Multiple regression model: N = a 0 + a 1 B 1 + a 2 B 2 + …… + a i B i + d;

[0117] M is or a and b are regression coefficients; N is the concentration value of the water quality index, a0, a1, …… ai are coefficients, d is the error term, and B1, B2, …… Bi are the reflectance values of the hyperspectral bands.

[0118] In S4, the method for establishing the quantitative mathematical model further includes an artificial neural network model, and its specific steps are as follows:

[0119] X1. Randomly select 60% of the data, use the reflectance of each band as the input layer and the measured water quality values for network training to construct a stable network;

[0120] X2. Then use the remaining 40% of the data to verify the accuracy of the model;

[0121] X3. By adjusting the weights and optimizing the model structure until the accuracy meets the requirements, the training ends and a stable model is formed;

[0122] The establishment of multiple quantitative mathematical models makes the results of data processing more accurate.

[0123] In S4, the calibration model is for calibrating the water quality remote sensing inversion model;

[0124] The calibration of the water quality remote sensing inversion model is used to test the practicability of the model, making the data analysis of the model more accurate;

[0125] The calibration of the water quality remote sensing inversion model is evaluated using the following formula:

[0126]

[0127]

[0128] where RMSE in the formula represents the root mean square error, MAPE represents the mean percentage error, and respectively represent the measured value and the predicted value of the water quality index at point i, and n represents the number of samples.

[0129] In S5, the steps of water quality remote sensing inversion and data visualization are as follows:

[0130] Z1. Use the data model established in S4 and import the model into the calculation tool;

[0131] Z2. Follow the prompts in the tool window to calculate step by step and obtain water quality hyperspectral remote sensing inversion data;

[0132] Z3. Classify the water quality hyperspectral remote sensing inversion data according to the data step level, assign different colors to each level, and save the output;

[0133] The data is generated as a visual map with different colors indicating the range of water quality concentration and spatial distribution characteristics; this map can be superimposed on the geographic base map to generate a remote sensing result map containing additional external geographical environment characteristics of the water area, making the data more intuitive. For example, the data obtained by the model produces a visual map with different colors indicating the range of water quality concentration and spatial distribution characteristics and superimposed on the geographic base map to generate a remote sensing result map containing additional external geographical environment characteristics of the water area. Figure 4 shown.

[0134] In S6, the extraction of the concentration anomaly point is obtained by analyzing the visualization data of the concentration distribution of each indicator. The morphological characteristics of the gradient distribution of the concentration are depicted with the anomaly point as the center point. This characteristic is the concentration diffusion characteristic of the pollutant. The diffusion of the pollutant concentration is manifested as a gradient decrease from the sewage outlet to the surrounding area. The fluid characteristics are generally point-shaped, block-shaped, diffuse, or band-shaped or line-shaped diffusion along the water flow direction; so that the concentration diffusion of the pollutant can be more clearly understood. For example, a typical fluid form diagram of the pollution concentration diffusion is shown in the following figure: Figure 5 shown.

[0135] In S7, the confirmation of the sewage discharge location refers to the point with the highest concentration according to the pollutant diffusion characteristics, that is, the location of the sewage outlet. According to the geographical location information provided by the image data, the longitude and latitude of the sewage outlet location are recorded; at the same time, the water quality index of the polluted discharge can be clearly indicated, that is, the type of pollution discharged can be determined, so that the degree of water pollution can be quickly and accurately surveyed and the pollution outlet can be found. For example, the schematic diagram of the remote sensing detection of pollutant emissions by drones is shown in Figure 6 shown.

[0136] It can conduct a rapid and accurate survey of the eutrophication and pollution level of water bodies, accurately locate pollution outlets, and have a more intuitive understanding of the spread of pollutant concentrations, allowing for timely water quality environmental monitoring and targeted prevention and control.

[0137] It should be noted that, in this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0138] Finally, it should be noted that the above are only preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An unmanned aerial vehicle (UAV) hyperspectral remote sensing method for detecting water pollution sources, characterized in that: The UAV hyperspectral remote sensing detection method includes the following steps: S1. Use UAV flight to obtain hyperspectral image data of the water body and synchronously collect ground water quality data; S2. Preprocess the obtained image data; S3. Extract spectral information of corresponding ground measured points according to the results of image data preprocessing; S4. Establish a quantitative mathematical model of water quality and hyperspectral data and verify the model; S5. Use the model to realize remote sensing inversion of water quality and visualization of concentration spatial distribution data; S6. Extract concentration anomaly points and conduct visual analysis on the fluid characteristics of concentration diffusion distribution; S7. Confirm the pollution type and pollution location; In S2, the preprocessing includes geometric correction, radiometric calibration, image mosaicking, and water area extraction; The geometric correction refers to obtaining a high-resolution satellite geographic image with a geographic coordinate system in the area where the target water area is located and which can be publicly obtained as a reference image for geometric correction, and using the geometric correction tool of remote sensing software to perform "image-image" geometric correction by the "homologous point matching" method; The radiometric calibration refers to finding the data area of the reference target on the UAV remote sensing image, extracting the reflectance data of more than 3 target reference plates, and then taking the average value; The calculation formula for the remote sensing reflectance data of the target reference plate is: ； Then, calibrate the spectral remote sensing DN value data of each pixel by the reflectance-based method to obtain the UAV spectral remote sensing reflectance data. The calculation formula of the reflectance-based method is: ； represents the calculated UAV remote sensing reflectance data ( ), represents the reflectance value of the target reference board obtained by actual measurement, represents the remote sensing reflectance value of the target reference board obtained by calculation ( , represents the spectral signal value measured by the UAV, represents the spectral signal value of the target reference board detected by the UAV; The image mosaicking refers to automatically mosaicking using the mosaicking tool of remote sensing software and manually polygon clipping using the clipping tool of remote sensing software; The water area extraction refers to extraction by setting a threshold according to the water body characteristic band. The water body threshold calculation formula is: NDVI = (B1 - B2) / (B1 + B2); NDWI = (B3 - B4) / (B3 + B4); where B1 and B3 are both the reflectances of the red light band, B2 and B4 are the reflectances of the near-infrared light band, NDVI is the normalized difference vegetation index, and NDWI is the normalized difference water index; Then extract the calculated threshold, generate a mask, and then convert the mask into a vector file of the water-land boundary through the ROI tool. Finally, use the ROI tool through the vector file to realize the clipping and extraction of the water body and obtain the hyperspectral image data of the water area.

2. The unmanned aerial vehicle (UAV) hyperspectral remote sensing method for detecting water pollution sources according to claim 1, characterized in that: In S1, the use of UAV flight to obtain hyperspectral image data of the water body refers to using UAV hyperspectral imaging remote sensing technology to obtain hyperspectral remote sensing image data of the target water area; The ground water quality data collection refers to selecting representative sample points, obtaining water samples through manual sampling, and bringing them back to the laboratory to detect the concentration of water quality indicators according to the analysis method of the Surface Water Environment Quality Standard GB 3838-2002.

3. The unmanned aerial vehicle (UAV) hyperspectral remote sensing method for detecting water pollution sources according to claim 1, characterized in that: In S3, the steps for extracting spectral information of corresponding ground measured points are: L1. Input the longitude and latitude of the actual measured points of surface water quality into a text in TXT format to generate a site longitude and latitude file; L2. Open the target image data and use the ROI from ASCII file tool to gradually import the corresponding files; L3. Then click the region of interest tool and select export to CSV to output the results in CSV format.

4. A method for drone hyperspectral remote sensing investigation of water pollution sources according to claim 1, characterized in that: In S4, the establishment of a quantitative mathematical model between water quality and hyperspectral data refers to using the water quality data collected in situ and the drone hyperspectral data for mathematical modeling to establish a quantitative relationship between water quality and hyperspectral data; The method for establishing the quantitative mathematical model between water quality and hyperspectral data is as follows: Establish empirical and semi-empirical models, including linear correlation models, polynomial models, and multiple regression models; The model establishment formula is as follows: Line-related model: Polynomial-related models: ; Multiple regression model: ; M is , or ; a and b are regression coefficients; N is the concentration value of the water quality index, a0, a1, …… ai are coefficients, d is the error term, and B1, B2, …… Bi are the reflectance values of the hyperspectral bands.

5. A method for drone hyperspectral remote sensing investigation of water pollution sources according to claim 1, characterized in that: In S4, the method for establishing the quantitative mathematical model between water quality and hyperspectral data also includes an artificial neural network model, and the specific steps are as follows: X1. Randomly select 60% of the data, use the reflectance of each band as the input layer and the actual measured water quality values for network training to construct a stable network; X2. Then use the remaining 40% of the data to verify the accuracy of the model; X3. By adjusting the weights and optimizing the model structure until the accuracy meets the requirements, the training ends and a stable model is formed.

6. A method for drone hyperspectral remote sensing investigation of water pollution sources according to claim 1, characterized in that: In S4, the verification of the model is to verify the water quality remote sensing inversion model; The water quality remote sensing inversion model is evaluated using the following formula: ； ； where RMSE represents the root mean square error and MAPE represents the mean absolute percentage error, and represent the measured value and the predicted value of the water quality index at point i, respectively, represents the number of samples.

7. A method for drone hyperspectral remote sensing investigation of water pollution sources according to claim 1, characterized in that: In S5, the steps for remote sensing inversion of water quality and visualization of concentration spatial distribution data are as follows: Z1. Use the data model established in S4 to import the model into the calculation tool; Z2. Calculate step by step according to the window prompts of the tool to obtain the hyperspectral remote sensing inversion data of water quality; Z3. Classify the hyperspectral remote sensing inversion data of water quality according to the data step level, assign different colors to each level, and save the output; The data generates a visualization map with different colors indicating the water quality concentration range and spatial distribution characteristics; this map is overlaid with a geographic base map to generate a remote sensing result map containing additional external geographic environment characteristics of the water area.

8. A method for drone hyperspectral remote sensing investigation of water pollution sources according to claim 1, characterized in that: In S6, the extraction of concentration anomaly points is obtained by analyzing the visualization data of the concentration distribution of each index. Taking the anomaly point as the center point, the morphological characteristics of the concentration gradient distribution are depicted, and this characteristic is the concentration diffusion characteristic of the pollutant.

9. A method for drone hyperspectral remote sensing investigation of water pollution sources according to claim 1, characterized in that: In S7, the confirmation of the sewage discharge type and location means that according to the pollutant diffusion characteristics, the point with the highest concentration is the sewage outlet location, and based on the geographical location information provided by the image data, the longitude and latitude of the sewage outlet location are recorded; at the same time, the water quality indicators of the pollution discharge are indicated, that is, the type of pollution discharged is determined.

Images