A small and micro water body water quality inversion method and system based on unmanned aerial vehicle spectral image

Abstract

The application relates to the technical field of water quality monitoring, and discloses a small and micro water body water quality inversion method and system based on an unmanned aerial vehicle spectral image, which comprises the following steps: uniformly arranging water body sampling points in a target water area, and recording position information of each sampling point; collecting water samples at each sampling point to obtain corresponding sampling water bodies, obtaining hyperspectral remote sensing images of the target water area, and pretreating the hyperspectral remote sensing images; obtaining spectral parameters V i of the hyperspectral remote sensing images and water quality information of the sampling water bodies, performing correlation analysis on the spectral parameters V i and the water quality information, training a water quality index inversion model, optimally selecting the best inversion model of each water quality index after model accuracy evaluation, obtaining a water quality spatial distribution map, analyzing the inversion result, and indicating that a pollution source is poured into a region with color mutation and pointing out possible pollution causes.

Description

Technical Field

[0001] This invention relates to the field of water quality monitoring technology, specifically to a method and system for water quality inversion of small water bodies based on UAV spectral imagery. Background Technology

[0002] Rural small and medium-sized rivers, as an important component of water resources, play a crucial role in irrigation, flood control, and environmental protection. However, rural rivers face threats from excessive use of pesticides and fertilizers, domestic sewage discharge, and untreated livestock and poultry waste runoff during the rainy season, increasing the risk of water quality deterioration.

[0003] Water quality monitoring is a crucial measure for protecting river water environments and provides a foundation for the scientific management of river water quality. The purpose of monitoring is to comprehensively understand the current status, spatiotemporal changes, and future trends of river water quality. Understanding water quality changes helps identify pollution sources, assess pollution intensity and impact range; therefore, rapid, real-time, accurate, and economical monitoring methods are needed.

[0004] Traditional water quality monitoring methods offer high accuracy but are time-consuming and labor-intensive. Furthermore, the measured data only represents information from local sampling points, making it difficult to quickly reflect the overall spatial water quality status and meet the requirements for real-time dynamic water quality monitoring. While satellite remote sensing technology has achieved a series of successes in the field of water quality monitoring, it is currently limited to large-scale water bodies such as rivers, lakes, and oceans, and cannot meet the water quality inversion needs of small, narrow waterways. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for water quality inversion of small water bodies based on UAV spectral imagery, and to solve the following technical problems:

[0006] How to acquire high spatial resolution images of small rivers to achieve accurate inversion of water quality parameters.

[0007] The objective of this invention can be achieved through the following technical solutions:

[0008] A method for water quality inversion of small water bodies based on UAV spectral imagery includes:

[0009] S1. Distribute water sampling points evenly in the target water area and record the location information of each sampling point;

[0010] S2. While collecting water samples at each sampling point to obtain the corresponding sampled water body, acquire hyperspectral remote sensing images of the target water area and preprocess the hyperspectral remote sensing images.

[0011] S3. Obtain the spectral parameters V of the hyperspectral remote sensing image. i And the water quality information of the sampled water body, for the spectral parameter V iCorrelation analysis was performed with water quality information;

[0012] Wherein, the spectral parameter V i The water quality index information is obtained based on the reflectance information of each band of the hyperspectral remote sensing image. The water quality index information includes multiple water quality indexes and their corresponding concentration values.

[0013] S4. Select the spectral parameter V that ranks among the top m in terms of correlation with each of the aforementioned water quality indicators. i As input to the water quality index inversion model, it is divided into a training set and a test set according to a specified ratio, and the output value of the inversion model is the concentration value of the corresponding water quality index.

[0014] S5. After evaluating the accuracy of the model, the best inversion model for each water quality index is selected to obtain the spatial distribution map of water quality. The inversion results are analyzed. Areas with abrupt color changes indicate the presence of pollution sources and the possible causes of pollution are pointed out.

[0015] As a further aspect of the present invention: step S1 includes:

[0016] S11. Acquire satellite images of the target water area, and distribute the sampling points evenly according to the area and shape of the target water area;

[0017] S12. Determine the latitude and longitude location information of each sampling point.

[0018] As a further aspect of the present invention: step S2 includes:

[0019] S21. After acquiring the target water area using a drone equipped with a hyperspectral camera, while collecting water samples, acquire the hyperspectral remote sensing images of the corresponding sampling points at the same time and under the same lighting conditions.

[0020] S22. Preprocess the hyperspectral remote sensing image;

[0021] The preprocessing includes mirror transformation, reflectivity calibration, and atmospheric correction.

[0022] As a further aspect of the present invention: step S3 includes:

[0023] S31. Extract the reflectance information of each band corresponding to each sampling point in the hyperspectral remote sensing image;

[0024] S32. Combine the reflectance values ​​of each single band into a band combination, and denot the result of the band combination as the spectral parameter V. i ;

[0025] S33, Regarding spectral parameter V i Correlation analysis was performed with water quality information.

[0026] As a further aspect of the present invention: the correlation analysis includes:

[0027]

[0028] Where k is the total number of sampling points, R xy x is the correlation coefficient between spectral parameters and water quality information. c With y c These are sample values ​​for spectral parameters and water quality information, respectively. and These represent the sample average values ​​of spectral parameters and water quality information, respectively.

[0029] As a further aspect of the present invention: in step S5, the method for selecting the optimal inversion model for each water quality index after model accuracy evaluation includes:

[0030] Based on the goodness of fit R of the inversion model 2 The model score S is obtained from the mean squared error (RMSE) and mean percentage error (MAPE).

[0031]

[0032] Where α and β are both preset weighting coefficients.

[0033] As a further aspect of the present invention: the model accuracy evaluation includes:

[0034] The goodness of fit R 2 How to obtain:

[0035]

[0036] The method for obtaining the mean squared error (RMSE) is as follows:

[0037]

[0038] The method for obtaining the mean percentage error (MAPE):

[0039]

[0040] Where n is the total number of validation samples, y i f is the actual value. i These are predicted values.

[0041] As a further aspect of the present invention: a water quality inversion system for small water bodies based on UAV spectral imagery, comprising:

[0042] The data acquisition module is used to uniformly deploy water sampling points in the target water area and record the location information of each sampling point; while collecting water samples at each sampling point to obtain the corresponding sampled water body, it acquires hyperspectral remote sensing images of the target water area and preprocesses the hyperspectral remote sensing images;

[0043] The feature extraction and analysis module is used to obtain the spectral parameters V of the hyperspectral remote sensing image. i And the water quality information of the sampled water body, for the spectral parameter V i Correlation analysis was performed with water quality information;

[0044] Wherein, the spectral parameter V i The water quality index information is obtained based on the reflectance information of each band of the hyperspectral remote sensing image. The water quality index information includes multiple water quality indexes and their corresponding concentration values.

[0045] The model building module is used to select the spectral parameters V that rank among the top m in terms of correlation with each of the water quality indicators. i As input to the water quality index inversion model, it is divided into a training set and a test set according to a specified ratio, and the output value of the inversion model is the concentration value of the corresponding water quality index.

[0046] The application and analysis module is used to select the best inversion model for each water quality index after the model accuracy is evaluated, obtain the spatial distribution map of water quality, analyze the inversion results, and indicate the presence of pollution sources in areas with abrupt color changes, pointing out the possible causes of pollution.

[0047] The beneficial effects of this invention are as follows: This invention provides a method and system for water quality inversion of small water bodies based on UAV hyperspectral imagery. During the detection process, water quality index concentration values ​​are obtained through physical and chemical detection methods. Water quality monitoring is achieved using UAV remote sensing, enabling the collection of measured data at sampling points in narrow rivers with complex water composition. The characteristic spectral parameters obtained from single-band / band combination are used to fit and train a model through traditional regression / machine learning algorithms. The inverted water quality status is then analyzed according to relevant standards and specifications, enabling the identification of water pollution levels and preliminary diagnosis of the possible causes of severe pollution in river sections, providing valuable data support for subsequent remediation. Attached Figure Description

[0048] The invention will now be further described with reference to the accompanying drawings.

[0049] Figure 1 This is a schematic diagram illustrating the principle framework of water quality inversion in this invention. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] Please see Figure 1 As shown, this invention is a method for water quality inversion of small water bodies based on UAV spectral imagery, comprising:

[0052] S1. Distribute water sampling points evenly in the target water area and record the location information of each sampling point;

[0053] S2. While collecting water samples at each sampling point to obtain the corresponding sampled water body, acquire hyperspectral remote sensing images of the target water area and preprocess the hyperspectral remote sensing images.

[0054] S3. Obtain the spectral parameters V of the hyperspectral remote sensing image. i And the water quality information of the sampled water body, for the spectral parameter V i Correlation analysis was performed with water quality information;

[0055] Wherein, the spectral parameter V i The water quality index information is obtained based on the reflectance information of each band of the hyperspectral remote sensing image. The water quality index information includes multiple water quality indexes and their corresponding concentration values.

[0056] S4. Select the spectral parameter V that ranks among the top m in terms of correlation with each of the aforementioned water quality indicators. i As input to the water quality index inversion model, it is divided into a training set and a test set according to a specified ratio, and the output value of the inversion model is the concentration value of the corresponding water quality index.

[0057] S5. After evaluating the accuracy of the model, the best inversion model for each water quality index is selected to obtain the spatial distribution map of water quality. The inversion results are analyzed. Areas with abrupt color changes indicate the presence of pollution sources and the possible causes of pollution are pointed out.

[0058] In this embodiment of the invention, m. can be selected as 5.

[0059] In step S1:

[0060] S11. Acquire satellite images of the target water area, and distribute the sampling points evenly according to the area and shape of the target water area;

[0061] S12. Determine the latitude and longitude location information of each sampling point.

[0062] In step S2, the flight path planning of the UAV can be driven by the accompanying software for experimental data collection.

[0063] S21. Use a drone equipped with a hyperspectral camera to acquire hyperspectral remote sensing images of the target water area. Compared with traditional manual sampling and monitoring and automatic monitoring station monitoring, drone monitoring is faster, has a wider range, and is more effective. Compared with satellite remote sensing monitoring, drone monitoring has higher resolution. While collecting water samples, operate the drone equipped with a GaiaSky-mini3-VN airborne spectral imager to take pictures over the water area to acquire the hyperspectral remote sensing images of the corresponding sampling points at the same time and under the same lighting conditions. The selected time period is sunny with a light breeze (level 1-2) and no cloud cover.

[0064] During water sampling, 500 mL of water sample was collected at each sampling point using a glass sampler at a depth of 0.5 m below the water surface. When collecting water samples, care was taken to avoid extremely narrow river channels and areas obstructed by floating debris such as duckweed. The collected water samples were promptly sent to the relevant laboratory for testing to obtain the concentration values ​​of the water quality indicators that needed to be retrieved. The water quality indicators measured included total nitrogen, total phosphorus, ammonia nitrogen, nitrate nitrogen, turbidity, and dissolved oxygen.

[0065] S22. Preprocess the hyperspectral remote sensing image, mainly including mirror transformation, reflectivity calibration, and atmospheric correction. The above processing is mainly completed using software such as SpectraVIEW.

[0066] In step S3:

[0067] S31. Extract the reflectance information of each band corresponding to each sampling point in the hyperspectral remote sensing image;

[0068] In this embodiment, reflectance extraction is performed as follows: A large amount of data is used to stitch together images of the study area. The stitched image data is exported as a .tiff format. ArcMap 10.8 software is opened, and the latitude and longitude corresponding to each sampling point of the ground water quality data are added to the layer as .xlsx / .xls data as input point features. The stitched UAV hyperspectral image is added to the layer as .tiff format as input raster data. In the toolbar, select System Toolbox - SpatialAnalyst Tools - Extraction Analysis - Multivalue Extraction to Points to open the attribute table, where the spectral reflectance values ​​of each band corresponding to the sampling points of the ground water quality data can be viewed.

[0069] S32. Combine the reflectance values ​​of each single band into a band combination, and denot the result of the band combination as the spectral parameter V. i Band combination operations, such as band ratios, addition, and subtraction, can reduce background noise, increase the sensitivity of spectral reflectance to water quality parameters, and improve inversion accuracy.

[0070] The main band combinations used in this study are as follows: R a +R b R a -R b R a / R b R a -R b / R a +R b ;

[0071] Among them, R a R b These represent the reflectance values ​​of any two bands within the 400–1000 nm wavelength range, with a ≠ b.

[0072] S33, Regarding spectral parameter V i Correlation analysis can be performed with water quality information. Specifically, Preson correlation analysis can be performed in SPSS software to calculate the correlation between spectral parameter Vi and the concentration values ​​of water quality indicators (total nitrogen, total phosphorus and ammonia nitrogen, nitrate nitrogen, turbidity, and dissolved oxygen).

[0073] The correlation analysis includes:

[0074]

[0075] Where k is the total number of sampling points, R xy x is the correlation coefficient between spectral parameters and water quality information. c With y c These are sample values ​​for spectral parameters and water quality information, respectively. and These are the sample average values ​​of spectral parameters and water quality information, respectively; R xy The closer the absolute value is to 1, the higher the correlation.

[0076] In step S4:

[0077] Select the top 5 spectral parameters V that are most correlated with various water quality indicators. i As the input to the inversion model for each water quality index, the output value is the concentration value of the corresponding water quality index;

[0078] The dataset is divided into training and test sets in a 7:3 ratio.

[0079] Training set data is used to train various traditional regression / machine learning models; validation set data is used in conjunction with various accuracy evaluation metrics (R²). 2 The water quality inversion model was evaluated using RMSE and MAPE.

[0080] The five band spectral parameters with the best correlation were used to divide the dataset into two parts: 70% for the training set and 30% for the validation set.

[0081] The traditional regression models in the training set mainly include linear models, exponential models, and multinomial models, while the machine learning models mainly include partial least squares (PLS), support vector machine (SVM), random forest (RF), and back propagation neural network (BP).

[0082] In step S5:

[0083] The method for selecting the optimal inversion model for each water quality index after model accuracy evaluation includes:

[0084] Based on the goodness of fit R of the inversion model 2 The mean squared error (RMSE) and mean percentage error (MAPE) are used to obtain the model score S; the higher the S, the better the inversion model.

[0085]

[0086] Where α and β are both preset weighting coefficients.

[0087] Specifically, the model accuracy evaluation includes:

[0088] The goodness of fit R 2 How to obtain:

[0089]

[0090] The method for obtaining the mean squared error (RMSE) is as follows:

[0091]

[0092] The method for obtaining the mean percentage error (MAPE):

[0093]

[0094] Where n is the total number of validation samples, y i f is the actual value. i These are predicted values.

[0095] As a further aspect of the present invention: a water quality inversion system for small water bodies based on UAV spectral imagery, comprising:

[0096] The data acquisition module is used to uniformly deploy water sampling points in the target water area and record the location information of each sampling point; while collecting water samples at each sampling point to obtain the corresponding sampled water body, it acquires hyperspectral remote sensing images of the target water area and preprocesses the hyperspectral remote sensing images;

[0097] The feature extraction and analysis module is used to obtain the spectral parameters V of the hyperspectral remote sensing image. i And the water quality information of the sampled water body, for the spectral parameter V i Correlation analysis was performed with water quality information;

[0098] Wherein, the spectral parameter V i The water quality index information is obtained based on the reflectance information of each band of the hyperspectral remote sensing image. The water quality index information includes multiple water quality indexes and their corresponding concentration values.

[0099] The model building module is used to select the spectral parameters V that rank among the top m in terms of correlation with each of the water quality indicators. i As input to the water quality index inversion model, it is divided into a training set and a test set according to a specified ratio, and the output value of the inversion model is the concentration value of the corresponding water quality index.

[0100] The application and analysis module is used to select the best inversion model for each water quality index after the model accuracy is evaluated, obtain the spatial distribution map of water quality, analyze the inversion results, and indicate the presence of pollution sources in areas with abrupt color changes, pointing out the possible causes of pollution.

[0101] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A method for water quality inversion of small water bodies based on UAV spectral imagery, characterized in that, include: S1. Acquire satellite images of the target water area, and uniformly distribute water sampling points in the target water area according to the area and shape of the target water area, and determine the latitude and longitude location information of each sampling point; S2. Using a drone equipped with a hyperspectral camera, water samples are collected at each sampling point to obtain the corresponding sampled water body. At the same time, hyperspectral remote sensing images of the target water body are acquired under the same lighting conditions at the same time. The hyperspectral remote sensing images are preprocessed, including mirror transformation, reflectivity calibration, and atmospheric correction. S3. Extract the reflectance information of each band corresponding to each sampling point in the hyperspectral remote sensing image, combine the reflectance values ​​of each single band, and record the result of the band combination as the spectral parameter. For spectral parameters Correlation analysis was performed between the water quality information of the sampled water body and the water quality information of the sampled water body; Wherein, the spectral parameters The band combinations are obtained based on the reflectance information of each band of the hyperspectral remote sensing image. + , - , , ,in, , These represent the reflectance values ​​of any two bands within the 400~1000nm wavelength range, and Water quality information includes multiple water quality indicators and their corresponding concentration values. The correlation analysis includes: ； in, The total number of sampling points. The correlation coefficient between spectral parameters and water quality information. and These are sample values ​​for spectral parameters and water quality information, respectively. These are the sample average values ​​of spectral parameters and water quality information, respectively. S4. Select the spectral parameters whose correlation with each of the aforementioned water quality indicators ranks in the top m. As input to the water quality index inversion model, the data is divided into a training set and a test set in a 7:3 ratio. The output value of the inversion model is the concentration value of the corresponding water quality index. The inversion model includes a traditional regression model and a machine learning model. The traditional regression model includes linear models, exponential models, and multinomial models. The machine learning model includes partial least squares models, support vector machine models, random forest models, and BP neural network models. S5. After evaluating the accuracy of the model, the best inversion model for each water quality index is selected to obtain the spatial distribution map of water quality. The inversion results are analyzed. Areas with abrupt color changes indicate the presence of pollution sources and the possible causes of pollution are pointed out. In step S5, the method for selecting the optimal inversion model for each water quality index after model accuracy evaluation includes: Based on the goodness of fit of the inversion model Mean square error and average percentage error Obtain model score ; ； in, and All are preset weighting coefficients; The goodness of fit How to obtain: ； The mean square error How to obtain: ； The average percentage error How to obtain: ； in, To verify the total number of samples, This is the actual value. These are predicted values.

2. A water quality inversion system for small water bodies based on UAV spectral imagery, the system being used to implement the water quality inversion method for small water bodies based on UAV spectral imagery as described in claim 1, characterized in that, include: The data acquisition module is used to uniformly deploy water sampling points in the target water area and record the location information of each sampling point; While collecting water samples at each sampling point to obtain the corresponding sampled water body, hyperspectral remote sensing images of the target water area are acquired, and the hyperspectral remote sensing images are preprocessed. The feature extraction and analysis module is used to obtain the spectral parameters of the hyperspectral remote sensing image. And the water quality information of the sampled water body, for spectral parameters Correlation analysis was performed with water quality information; Wherein, the spectral parameters The water quality index information is obtained based on the reflectance information of each band of the hyperspectral remote sensing image. The water quality index information includes multiple water quality indexes and their corresponding concentration values. The model building module is used to select the spectral parameters whose correlation with each of the water quality indicators ranks in the top m. As input to the water quality index inversion model, it is divided into a training set and a test set according to a specified ratio, and the output value of the inversion model is the concentration value of the corresponding water quality index. The application and analysis module is used to select the best inversion model for each water quality index after the model accuracy is evaluated, obtain the spatial distribution map of water quality, analyze the inversion results, and indicate the presence of pollution sources in areas with abrupt color changes, pointing out the possible causes of pollution.

Images