River water quality inversion and health evaluation method and system based on satellite remote sensing, terminal equipment and medium

Abstract

The invention discloses a river water quality inversion and health evaluation method and system based on satellite remote sensing, terminal equipment and a medium, and relates to the technical field of environment remote sensing monitoring. The method comprises the following steps: acquiring a time-space matched multispectral remote sensing image and ground actually measured water quality parameter data set of a research area; preprocessing the image and extracting a water body area image; constructing a data sample set based on the water body area image and the water quality parameter data set; establishing a plurality of inversion models for each water quality parameter, and selecting an optimal inversion model through performance verification; and calculating a comprehensive health index and a health level of the river in the water body area through objective weight determination. According to the method, the problems of low monitoring efficiency, insufficient inversion precision and high evaluation subjectivity in the prior art are solved, efficient and accurate inversion and objective health evaluation of large-range river water quality are realized, and decision support is provided for environment management.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of environmental remote sensing monitoring, and in particular to a river water quality inversion and health evaluation method and system based on satellite remote sensing, a terminal device and a medium. BACKGROUND

[0002] River water quality monitoring is the core work of environmental protection. Traditional monitoring methods rely on ground fixed-point sampling, which has inherent defects such as high cost, low efficiency, and difficulty in achieving large-scale synchronous monitoring. Although remote sensing technology can provide a macro perspective, existing solutions still have obvious bottlenecks. First, the inversion accuracy is insufficient, and traditional statistical models or single algorithms are mostly used, which have limited adaptability and accuracy for complex water bodies. Second, the evaluation is highly subjective, and the weights in the comprehensive health evaluation link mostly rely on expert experience or simple averaging, lacking objectivity and being difficult to standardize and promote. Third, the process is fragmented, and the water quality inversion and health evaluation links are independent of each other, without forming an integrated automatic processing process, making it difficult to serve efficient business operation.

[0003] Therefore, there is an urgent need for an integrated method that can achieve large-scale river water quality inversion with high precision and objective health evaluation to fill the gap in existing technology. SUMMARY

[0004] The technical problem to be solved by the present application is that in the field of environmental remote sensing monitoring, the existing river water quality monitoring method has the problems of difficulty in large-scale synchronous monitoring, low water quality inversion accuracy, strong subjectivity in health evaluation, and fragmented process, which cannot meet the needs of efficient, accurate, and objective business monitoring. Therefore, there is an urgent need for an effective solution to solve the above technical problems.

[0005] To solve the above technical problems, the technical solution adopted by the present application is as follows: In a first aspect, the present application provides a river water quality inversion and health evaluation method based on satellite remote sensing, which comprises: obtaining multispectral remote sensing images of a study area, and obtaining water quality parameter datasets through ground measurement; wherein the multispectral remote sensing images and the water quality parameter datasets are spatio-temporally matched; preprocessing the multispectral remote sensing images and extracting images of water body regions in the study area; based on the images of the water body regions and the corresponding water quality parameter datasets, constructing a data sample set; based on the data sample set, constructing several inversion models for each water quality parameter in the water quality parameter dataset, and obtaining the optimal inversion model corresponding to each water quality parameter through performance verification; determining the weights of each water quality parameter through an objective weighting method, and calculating the river health comprehensive index and the river health grade of the water body region.

[0006] In an implementation manner, the multispectral remote sensing image of the research area is acquired, and the water quality parameter dataset is obtained through ground measurement, including: downloading the multispectral remote sensing image of a specific time phase of the research area from a remote sensing satellite data platform; during the period when the multispectral remote sensing image passes through, ground water quality sampling is performed on a plurality of sampling points in the research area to obtain water quality parameters of each sampling point, and the position coordinates of each sampling point are recorded; integrating the water quality parameters and the corresponding position coordinates of all sampling points to form the water quality parameter dataset; wherein the water quality parameters include at least one of chlorophyll a, chemical oxygen demand, total phosphorus, and transparency.

[0007] In an implementation manner, the multispectral remote sensing image is preprocessed, and the image of the water body region in the research area is extracted, including: performing radiation calibration processing and atmospheric correction processing on the multispectral remote sensing image to eliminate radiation errors and atmospheric interference of the image, and acquiring the ground true reflectance data of the research area; based on the ground true reflectance data, using the normalized water body index to calculate the water body index image of the research area; based on a pre-set water body extraction threshold, performing pixel screening on the water body index image of the research area, retaining the pixels corresponding to the water body region and excluding the pixels corresponding to the non-water body region, to obtain the image of the water body region in the research area; wherein the image of the water body region is stored in the form of a water body mask.

[0008] In an implementation manner, the data sample set is constructed based on the image of the water body region and the corresponding water quality parameter dataset, including: based on the position coordinates of each sampling point in the water quality parameter dataset, locating the pixels corresponding to each sampling point in the image of the water body region, and extracting the spectral data of each corresponding pixel; wherein the spectral data includes the reflectance of each band of the multispectral remote sensing image, and the spectral index calculated from the reflectance of each band; associating and matching the water quality parameters of each sampling point with the spectral data of the corresponding pixel of the sampling point to form a plurality of groups of data pairs of water quality parameters and spectral data; summarizing all data pairs to obtain an initial dataset, and randomly dividing the initial dataset according to a pre-set proportion to obtain the data sample set composed of a training set and a validation set.

[0009] In an implementation manner, the constructing a plurality of inversion models for each water quality parameter in the water quality parameter dataset based on the data sample set and obtaining an optimal inversion model corresponding to each water quality parameter through performance verification comprises: For each water quality parameter in the water quality parameter dataset, at least two quantitative regression models are respectively constructed as candidate inversion models of the water quality parameter based on a training set in the data sample set; wherein the quantitative regression model comprises a random forest and a support vector regression; The spectral data of each pixel in the validation set in the data sample set is respectively input into each candidate inversion model, and a water quality parameter prediction value output by each candidate inversion model is calculated; Based on the measured value of the water quality parameter in the validation set in the data sample set and the water quality parameter prediction value output by each candidate inversion model, a determination coefficient, a root mean square error and a mean absolute error of each candidate inversion model are respectively calculated; The determination coefficient, the root mean square error and the mean absolute error of each candidate inversion model under the same water quality parameter are compared, and the candidate inversion model with the maximum determination coefficient and the minimum root mean square error and mean absolute error is selected as the optimal inversion model corresponding to the water quality parameter.

[0010] In an implementation manner, the determining the weight of each water quality parameter through an objective weighting method comprises: The water quality parameter of the water body region output by the optimal inversion model is integrated into a water quality data matrix; Each water quality parameter in the water quality data matrix is subjected to standardization processing, and the weight of each water quality parameter is calculated through an entropy weight method.

[0011] In an implementation manner, the calculating the river health comprehensive index and the river health grade of the water body region comprises: Based on the weight of each water quality parameter and the water quality parameter of the water body region, the river health comprehensive index of each pixel in the water body region is calculated; Based on the river health comprehensive index and a pre-set health grade division threshold, the health grade of each pixel in the water body region is divided to obtain the river health grade of each pixel in the water body region; The river health grades of each pixel in the water body region are integrated to obtain a river health grade spatial distribution map of the research region.

[0012] In a second aspect, the embodiment of the present application further provides a river water quality inversion and health evaluation system based on satellite remote sensing, which comprises: An original data acquisition module is configured to acquire multispectral remote sensing images of a study area and obtain a water quality parameter dataset through ground measurement; wherein the multispectral remote sensing images and the water quality parameter dataset are spatio-temporally matched; A data preprocessing module is configured to preprocess the multispectral remote sensing images and extract images of water body regions in the study area; A sample set construction module is configured to construct a data sample set based on the images of the water body regions and the corresponding water quality parameter dataset; A model selection and verification module is configured to construct several inversion models for each water quality parameter in the water quality parameter dataset based on the data sample set, and obtain an optimal inversion model corresponding to each water quality parameter through performance verification; A comprehensive evaluation module is configured to determine the weights of each water quality parameter through an objective weighting method, and calculate a river health comprehensive index and a river health grade of the water body region.

[0013] In a third aspect, an embodiment of the present application further provides a terminal device, which comprises a memory, a processor, and a satellite remote sensing based river water quality inversion and health evaluation program stored in the memory and executable on the processor; when the processor executes the satellite remote sensing based river water quality inversion and health evaluation program, the steps of the satellite remote sensing based river water quality inversion and health evaluation method in any of the above solutions are implemented.

[0014] In a fourth aspect, an embodiment of the present application further provides a computer readable storage medium, which stores a satellite remote sensing based river water quality inversion and health evaluation program; when a processor executes the satellite remote sensing based river water quality inversion and health evaluation program, the steps of the satellite remote sensing based river water quality inversion and health evaluation method in any of the above solutions are implemented.

[0015] Beneficial effects: The application discloses a river water quality inversion and health evaluation method and system based on satellite remote sensing, a terminal device and a medium, and relates to the technical field of environmental remote sensing monitoring. The method comprises the following steps: acquiring a multispectral remote sensing image of a research area, and obtaining a water quality parameter dataset through ground measurement; wherein the multispectral remote sensing image and the water quality parameter dataset are spatio-temporally matched. Then, the multispectral remote sensing image is preprocessed, and an image of a water body region in the research area is extracted. Next, based on the image of the water body region and the corresponding water quality parameter dataset, a data sample set is constructed. Subsequently, based on the data sample set, a plurality of inversion models are constructed for each water quality parameter in the water quality parameter dataset, and the optimal inversion model corresponding to each water quality parameter is obtained through performance verification. Finally, the weight of each water quality parameter is determined through an objective weighting method, and the river health comprehensive index and the river health grade of the water body region are calculated. The application optimizes the optimal inversion model through multi-model construction and performance verification, improves the river water quality inversion accuracy and robustness, replaces the subjective experience weighting with the objective weighting method, avoids human subjective bias, and makes the health evaluation result more fair and comparable. Through the whole process from data acquisition to health grade output, automatic processing is realized, the efficiency is high, and the business operation can be realized. At the same time, relying on the advantages of satellite remote sensing, large-scale, long-time dynamic monitoring can be realized, pollution hotspots can be accurately identified, and powerful decision support can be provided for environmental management. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 The flow chart of the specific implementation mode of the river water quality inversion and health evaluation method based on satellite remote sensing provided by the embodiment of the application is shown.

[0017] Figure 2 The detailed flow chart of the river water quality inversion and health evaluation method based on satellite remote sensing provided by the embodiment of the application is shown.

[0018] Figure 3 The distribution diagram of the health grade of a river basin generated by the river water quality inversion and health evaluation method based on satellite remote sensing provided by the embodiment of the application is shown.

[0019] Figure 4 The principle block diagram of the river water quality inversion and health evaluation device based on satellite remote sensing provided by the embodiment of the application is shown.

[0020] Figure 5 The internal structure principle block diagram of the terminal device provided by the embodiment of the application is shown. DETAILED DESCRIPTION

[0021] In order to make the objects, technical solutions and effects of the present application clearer and more explicit, the present application will be further described in detail below with reference to the drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application and do not limit the present application.

[0022] The flowchart shown in the drawings is only an example and does not necessarily include all contents and operations or steps, nor does it necessarily be executed in the order described. For example, some operations or steps can also be decomposed, combined or partially merged, so that the actual execution order can be changed according to the actual situation.

[0023] It should be understood that the terms used in the present application are only for the purpose of describing specific examples and are not intended to limit the present application. As used in the present application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include the plural forms.

[0024] It should be understood that, in order to facilitate the clear description of the technical solutions of the embodiments of the present application, in the embodiments of the present application, the terms "first", "second" and the like are used to distinguish the same or similar items with basically the same function and effect. For example, the first control information and the second control information are only used to distinguish different control information, and do not limit the order.

[0025] Those skilled in the art can understand that the terms "first", "second" and the like do not limit the quantity and execution order, and the terms "first", "second" and the like do not necessarily mean different.

[0026] It should also be understood that the term "and / or" used in the present application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0027] River water quality is an important indicator to maintain the ecological balance of the basin and ensure the safety of human production and living water. Precise and efficient river water quality monitoring and health evaluation is the key foundation for water environment management, pollution traceability treatment and ecological protection and restoration. In the traditional river water quality monitoring work, it mainly relies on ground fixed-point sampling combined with laboratory analysis. Although this method can obtain accurate water quality data at a single point, it has two defects. On the one hand, the sampling process requires a lot of manpower, material resources and time cost, and the sampling points are limited, which is difficult to realize synchronous monitoring of a large range of basin, and cannot fully reflect the spatial distribution difference and dynamic change characteristics of river water quality. On the other hand, the laboratory analysis period is long, which is difficult to quickly respond to sudden water pollution events, resulting in a lag in pollution control decision-making.

[0028] With the development of remote sensing technology, satellite remote sensing has gradually been applied to river water quality monitoring field with its advantages of macroscopic, rapid and periodic observation, providing a new technical approach for large-scale water environment monitoring. However, there are still many bottlenecks to be solved in the existing satellite remote sensing-based river water quality inversion and health evaluation technology, which is difficult to meet the actual business monitoring and management needs. Firstly, the precision of remote sensing image preprocessing is insufficient, and there is great uncertainty in the atmospheric correction link. Even the current mainstream atmospheric correction processor, the error of the key spectral band is still high in inland river water reflectance inversion, and this error will further conduct to the downstream water quality parameter inversion result, leading to high uncertainty of the inversion of core water quality parameters such as chlorophyll a and total suspended solids, which seriously affects the inversion precision. Secondly, the adaptability and stability of water quality inversion model are poor. The existing technology mostly uses single statistical model or machine learning algorithm to construct the inversion model, without comparing and optimizing multiple models according to the characteristics of different water quality parameters, and the model has insufficient adaptability to the complex and changeable river water optical properties, which is easy to cause large deviation of the inversion result. Thirdly, the health evaluation is highly subjective. In the existing evaluation method, the weight of each water quality parameter is determined mainly depending on the expert experience judgment or simple average weighting, lacking of quantitative weighting means based on objective data, which leads to that the evaluation result is significantly affected by human factors, and the results obtained by different evaluation subjects are difficult to compare, which cannot realize standardized promotion. Fourthly, the technical process is fragmented. In the existing technology, the remote sensing image acquisition, preprocessing, water quality inversion and health evaluation are independent of each other, which needs manual intervention for data connection and processing. Not only the efficiency is low, but also the manual operation is easy to introduce errors. Early, even the remote sensing image needs to be interpreted point by point manually, and it often takes three days to analyze one image, which is difficult to meet the needs of business efficient operation.

[0029] Therefore, there is an urgent need for a method for realizing high-precision inversion and objective health evaluation of large-scale river water quality, which can provide accurate and efficient decision support for water environment management.

[0030] The method for river water quality inversion and health evaluation based on satellite remote sensing provided in the embodiment is applied to a mixed reality device, such as Figure 1 As shown in the figure, the method specifically comprises the following steps: Step S100, acquiring a multi-spectral remote sensing image of a research area, and obtaining a water quality parameter data set through ground measurement; wherein the multi-spectral remote sensing image and the water quality parameter data set are spatio-temporally matched.

[0031] In the embodiment, the multispectral remote sensing image refers to remote sensing data containing multiple spectral band reflectivity information acquired by a satellite sensor, which can capture the reflection characteristics of water bodies to electromagnetic waves of different wavelengths. The ground measured water quality parameter dataset refers to a collection of water quality index data obtained by field sampling in the study area combined with laboratory analysis or on-site detection equipment. Specifically, it can include measured values of chlorophyll a (Chl-a), chemical oxygen demand (COD), total phosphorus (TP), transparency (SDD), etc. The space-time matching refers to the imaging time of the multispectral remote sensing image being consistent with the ground sampling time or being within a predetermined time range, for example, within 1 hour before and after the image passes, and the spatial position of the ground sampling point accurately corresponds to the image element position. The spatial correspondence is realized by recording the position coordinates of the sampling point, which can be obtained by Beidou positioning.

[0032] By acquiring the space-time matched double data sources, on the one hand, the multispectral remote sensing image is used to cover a large range of basins, solving the problem of limited coverage range of traditional ground fixed-point sampling. On the other hand, the ground measured data provides a real verification basis for subsequent water quality inversion model, avoiding the deviation of the inversion result from the actual situation caused by relying only on remote sensing data, and providing a data basis for subsequent high-precision water quality inversion.

[0033] In one implementation manner, the multispectral remote sensing image of the study area is acquired, and the water quality parameter dataset is obtained through ground measurement. Specifically, the following steps are included: Step S110, downloading the multispectral remote sensing image of a specific phase of the study area from a remote sensing satellite data platform; Step S120, during the period when the multispectral remote sensing image passes, ground water quality sampling is performed on a plurality of sampling points in the study area to obtain water quality parameters of each sampling point, and the position coordinates of each sampling point are recorded; Step S130, integrating the water quality parameters and corresponding position coordinates of all sampling points to form the water quality parameter dataset; The water quality parameters include at least one of chlorophyll a, chemical oxygen demand, total phosphorus, and transparency.

[0034] In the embodiment, as shown in Figure 2 First, the collection of remote sensing data and ground measured data is acquired to prepare the basic data.

[0035] When acquiring the multispectral remote sensing image of the study area, the multispectral remote sensing image of a specific phase of the study area is downloaded from a remote sensing satellite data platform. The specific phase is determined according to the research requirements, for example, images of key periods such as no cloud or little cloud, water quality monitoring such as flood season, dry season, etc.

[0036] When the water quality parameter dataset is obtained through ground measurement, a plurality of sampling points are arranged in the study area during the time period when the multispectral remote sensing image passes, and the water samples at each point are collected by using the standard water quality sampling method. Then, the water quality parameters of each sampling point are measured by using the laboratory analysis equipment. At the same time, the accurate position coordinates of each sampling point are recorded by using the Beidou positioning equipment, so that the spatial position of the sampling point can be accurately corresponded to the pixel of the remote sensing image.

[0037] Finally, the water quality parameters and the corresponding position coordinates of all sampling points are integrated to form a water quality parameter dataset, wherein the water quality parameters can specifically include at least one of chlorophyll-a, chemical oxygen demand, total phosphorus and transparency.

[0038] By ensuring the spatial and temporal matching accuracy of the multispectral remote sensing image and the water quality parameter dataset, the data correlation error caused by the time difference and the spatial deviation is reduced, and the basis for subsequent construction of a high-quality data sample set is provided.

[0039] Step S200, pre-processing the multispectral remote sensing image, and extracting the image of the water body region in the study area.

[0040] In the embodiment, the multispectral remote sensing image preprocessing refers to a data processing process for eliminating the errors generated in the acquisition process of the remote sensing image to obtain data reflecting the true reflection characteristics of the ground, which can specifically include radiation calibration, atmospheric correction and the like. The image of the water body region refers to the image data of the water body region including only rivers, lakes and the like separated from the pre-processed multispectral remote sensing image, and does not include land, vegetation, clouds and other non-water body regions.

[0041] In the preprocessing process, the radiation calibration can convert the digital quantization value of the image into a radiation brightness value with physical meaning, eliminate the errors caused by the response difference of the sensor itself, and the atmospheric correction can eliminate the influence of atmospheric scattering, absorption and the like on the spectral reflectance, to obtain the true reflectance data of the ground. Then, when the water body region image is extracted, the specific water body recognition method is focused on the target region to exclude the interference of the non-water body region, so that the subsequent water quality inversion is only carried out for the river water body, and the spectral information of the non-water body pixel is avoided to interfere with the inversion result, thereby improving the pertinence and accuracy of the inversion.

[0042] In one implementation manner, the pre-processing of the multispectral remote sensing image and the extraction of the image of the water body region in the study area specifically include the following steps: Step S210, performing radiation calibration processing and atmospheric correction processing on the multispectral remote sensing image to eliminate the radiation error and atmospheric interference of the image, and obtaining the true reflectance data of the ground in the study area; Step S220: Based on the actual surface reflectance data, calculate the water index image of the study area using the normalized water index; Step S230: Based on a preset water extraction threshold, perform pixel filtering on the water index image of the study area, retain the pixels corresponding to the water area and remove the pixels corresponding to the non-water area to obtain the image of the water area in the study area; wherein, the image of the water area is stored in the form of a water mask.

[0043] In this embodiment, as Figure 2 As shown, the basic data preprocessing is achieved through radiometric calibration and atmospheric correction, as well as the extraction of water areas.

[0044] When preprocessing multispectral remote sensing images, radiometric calibration is performed first, converting the digital quantization values ​​of the image into radiance values. This can be calculated using the radiometric calibration coefficient file provided with the satellite sensor. Following this, atmospheric correction is performed to eliminate the influence of atmospheric scattering, absorption, and aerosols on spectral reflectance. Specifically, the FLAASH atmospheric correction algorithm is used. Based on atmospheric conditions such as visibility and water vapor content at the time of image formation, an atmospheric radiative transfer model is established, and the true surface reflectance data is retrieved. The FLAASH atmospheric correction algorithm is a commonly used high-precision atmospheric correction algorithm in remote sensing image preprocessing.

[0045] When extracting images of water bodies, the water index image of the study area is calculated using the Normalized Difference Water Index (NDWI) based on the true surface reflectance data. Subsequently, based on a preset water extraction threshold, the water index image is filtered to retain pixels with a water index value greater than the threshold (i.e., water body area pixels) and remove pixels with a non-water body index value less than or equal to the threshold (i.e., land, vegetation, and other non-water body area pixels). Finally, the image of the water body area in the study area is obtained, and the image of the water body area is stored in the form of a water mask, in which the water body area pixel value is set to 1 and the non-water body area pixel value is set to 0.

[0046] Preprocessing effectively eliminates image errors and ensures the authenticity of spectral reflectance data. At the same time, by combining NDWI with threshold screening, water and non-water areas can be accurately separated, avoiding interference from non-water pixels in subsequent water quality inversion and improving the accuracy of water extraction.

[0047] Step S300: Construct a data sample set based on the image of the water body area and the corresponding water quality parameter dataset.

[0048] In the embodiment, the data sample set refers to a data set formed by associating the spectral information in the image of the water area with the ground measured water quality parameters, which contains the corresponding relationship between the spectral data and the water quality parameters, and is the basic data for constructing the water quality inversion model. The spectral data can include spectral reflectance values.

[0049] When constructing the data sample set, first, according to the spatial position information of each sampling point in the ground measured water quality parameter data set, the corresponding pixel in the image of the water area is located, and the spectral data of the pixel is matched with the water quality parameter of the sampling point. Subsequently, the water quality parameter of each sampling point, such as the measured value of chlorophyll a of a point, is associated with the spectral reflectance value of the corresponding pixel one by one to form a plurality of data pairs of water quality parameters and spectral data. Finally, all the data pairs are integrated to obtain a complete data sample set.

[0050] The sample set constructed in this way can directly establish the correlation between the water quality parameters and the spectral features, provide reliable training data for the subsequent inversion model, and avoid the problem of inaccurate model construction caused by data correlation deviation.

[0051] In an implementation manner, the data sample set is constructed based on the image of the water area and the corresponding water quality parameter data set, and specifically includes the following steps: Step S310, based on the position coordinates of each sampling point in the water quality parameter data set, locating the corresponding pixel in the image of the water area, and extracting the spectral data of each corresponding pixel; wherein the spectral data includes the reflectance of each band of the multispectral remote sensing image, and the spectral index calculated from the reflectance of each band; Step S320, associating and matching the water quality parameter of each sampling point with the spectral data of the corresponding pixel of the sampling point to form a plurality of data pairs of water quality parameters and spectral data; Step S330, collecting all the data pairs to obtain an initial data set, and randomly dividing the initial data set according to a preset proportion to obtain the data sample set composed of a training set and a verification set.

[0052] In the embodiment, as shown in Figure 2 In the model construction and verification stage, first, the matching of spectral and water quality data is performed.

[0053] When constructing the data sample set based on the image of the water area and the water quality parameter data set, first, based on the position coordinates of each sampling point in the water quality parameter data set, the corresponding pixels of each sampling point are located in the image of the water area through spatial overlay analysis, and then the spectral data of each corresponding pixel is extracted. Specifically, the spectral data includes the reflectance of each band of the multispectral remote sensing image, such as the reflectance of the blue band, the green band, the red band, and the near-infrared band, and the spectral index calculated from the reflectance of each band, such as NDVI (normalized difference vegetation index) and NDWI.

[0054] Subsequently, the water quality parameters of each sampling point are associated and matched with the spectral data of the corresponding pixel of the point, to form a plurality of data pairs of water quality parameters and spectral data. Then, all the data pairs are summarized to obtain an initial data set, and the initial data set is randomly divided according to a preset proportion, which can be 70% for the training set and 30% for the validation set. The training set is used for subsequent construction of the inversion model, and the validation set is used for verification of the performance of the model. The training set and the validation set together constitute the data sample set.

[0055] By extracting the spectral data containing band reflectance and spectral index, more feature information is provided for the inversion model, which helps to improve the prediction ability of the model for water quality parameters. At the same time, the division ratio of 70% for the training set and 30% for the validation set ensures sufficient data for training the model while effectively verifying the generalization ability of the model, avoiding overfitting or underfitting of the model.

[0056] Step S400, based on the data sample set, a plurality of inversion models are constructed for each water quality parameter in the water quality parameter data set, and the optimal inversion model corresponding to each water quality parameter is obtained through performance verification.

[0057] In this embodiment, the inversion model refers to a mathematical model that can establish a quantitative relationship between the water quality parameter and the spectral data. Through the model, the corresponding water quality parameter inversion value can be calculated according to the spectral data of the image. The plurality of inversion models refers to a plurality of inversion models constructed by using different algorithms for the same water quality parameter, which can specifically include models based on ensemble learning and regression models based on supervised learning. Performance verification refers to a process of evaluating the prediction effect of the inversion model by calculating a specific precision index using the preset validation data.

[0058] For each water quality parameter, such as chlorophyll a and chemical oxygen demand, a plurality of machine learning algorithms are used to construct quantitative regression models, and then the validation data in the data sample set is input into each model to calculate the precision index of the model, such as the coefficient of determination, the root mean square error, and the mean absolute error. Then, the precision indexes of the models are compared, and the model with the best precision is selected as the optimal inversion model for the water quality parameter.

[0059] Through multi-model construction and optimization, the advantages of different algorithms can be fully utilized, the problem of insufficient adaptability of a single model to complex water bodies can be avoided, and the precision and robustness of water quality parameter inversion can be effectively improved.

[0060] In an implementation manner, the plurality of inversion models are constructed for each water quality parameter in the water quality parameter dataset based on the data sample set, and the optimal inversion model corresponding to each water quality parameter is obtained through performance verification, and the method specifically comprises the following steps: In step S410, at least two quantitative regression models are respectively constructed as candidate inversion models of each water quality parameter in the water quality parameter dataset based on the training set in the data sample set; wherein the quantitative regression model comprises a random forest and a support vector regression. In step S420, the spectral data of each pixel in the validation set in the data sample set is input into each candidate inversion model, and the water quality parameter prediction value output by each candidate inversion model is calculated. In step S430, the determination coefficient, the root mean square error and the mean absolute error of each candidate inversion model are respectively calculated based on the measured value of the water quality parameter in the validation set in the data sample set and the water quality parameter prediction value output by each candidate inversion model. In step S440, the determination coefficient, the root mean square error and the mean absolute error of each candidate inversion model under the same water quality parameter are compared, and the candidate inversion model with the maximum determination coefficient and the minimum root mean square error and mean absolute error is selected as the optimal inversion model corresponding to the water quality parameter.

[0061] In the embodiment, as shown in FIG. 4, in the model construction and verification stage, regression analysis and model training are performed, and model optimization and verification are performed. Figure 2

[0062] When the inversion model is constructed based on the data sample set and the performance is verified, for each water quality parameter in the water quality parameter dataset, the measured value of the water quality parameter in the training set in the data sample set is used as the dependent variable, and the spectral data of the corresponding pixel in the training set is used as the independent variable, and two quantitative regression models are respectively constructed using a random forest algorithm and a support vector regression algorithm as the candidate inversion model of the water quality parameter. The random forest algorithm constructs multiple decision trees, each decision tree is trained based on the bootstrap sampling set of the training set, and only a part of features selected randomly is considered when the node of the tree is split, and finally the prediction results of all decision trees are averaged to obtain the model output. The support vector regression algorithm finds an interval band so that most of the training data points fall within the interval band, while maintaining the flatness of the interval band center function, to realize the regression prediction of the water quality parameter.

[0063] ​Subsequently, the spectral data of each pixel in the validation set of the data sample set were input into the two candidate inversion models for the water quality parameter, and the predicted water quality parameter output by each candidate inversion model was calculated. Based on the measured and predicted water quality parameter values ​​in the validation set, the coefficient of determination (R²), root mean square error (RMSE), and mean absolute error (MAE) of each candidate inversion model were calculated. The coefficient of determination measures the model's ability to explain the variation in water quality parameters, with a value ranging from [0,1]. The closer to 1, the higher the model accuracy. The root mean square error measures the average deviation between the predicted and measured values; the smaller the value, the higher the model accuracy. The mean absolute error measures the absolute deviation between the predicted and measured values; the smaller the value, the higher the model accuracy.

[0064] Finally, by comparing the coefficient of determination, root mean square error, and mean absolute error of the two candidate inversion models under the same water quality parameter, the candidate inversion model with the largest coefficient of determination and the smallest root mean square error and mean absolute error is selected as the optimal inversion model for that water quality parameter.

[0065] The root mean square error (RMSE) is expressed as follows:

[0066] The coefficient of determination (R²) is expressed as follows:

[0067] The expression for Mean Absolute Error (MAE) is:

[0068] in, The amount of water quality parameter sample data. The selected water quality parameter number is The sample predicted value, The selected water quality parameter number is The true value of the sample.

[0069] In this embodiment, experimental data is provided to show the selected optimal inversion model and its corresponding coefficient of determination, root mean square error, and mean absolute error, as shown in Table 1.

[0070] Table 1

[0071] As shown in Table 1, for chlorophyll a, the R² of the random forest model is 0.87, the RMSE is 2.45 μg / L, and the MAE is 1.82 μg / L, which is better than the support vector regression model, so the optimal inversion model of chlorophyll a is the random forest model; for chemical oxygen demand, the R² of the support vector regression model is 0.79, the RMSE is 1.05 mg / L, and the MAE is 0.78 mg / L, which is better than the random forest model, so the optimal inversion model of chemical oxygen demand is the support vector regression model. That is, it is verified that the optimal model of chlorophyll a and transparency is random forest (R²> 0.85), and the optimal model of chemical oxygen demand and total phosphorus is support vector machine (R²> 0.78).

[0072] Through this multi-model construction and optimization method, the advantages of different algorithms can be fully utilized, and the highest precision inversion model can be ensured for each water quality parameter, effectively improving the overall precision of water quality inversion.

[0073] Step S500, the weight of each water quality parameter is determined by an objective weighting method, and the river health comprehensive index and the river health level of the water body region are calculated.

[0074] In this embodiment, the objective weighting method refers to a method of determining the weight of each parameter according to the dispersion degree (i.e., data variation characteristics) of the water quality parameter data itself without relying on artificial subjective judgment, which can avoid the subjective bias caused by expert experience weighting. The river health comprehensive evaluation index (RHHI) refers to a value that can quantify the river health degree obtained by weighting and summing the standardized values of each water quality parameter and the corresponding weight. The river health level refers to the level category reflecting the river health condition divided according to the size of the river health comprehensive evaluation index.

[0075] First, the weight of each water quality parameter is calculated by the objective weighting method to ensure the fairness of weight distribution. Then, the water quality parameter inversion values of each pixel in the water body region are standardized to eliminate the influence of different parameter dimension differences. Subsequently, the river health comprehensive evaluation index of each pixel is calculated according to the weight and the standardized water quality parameter value. Finally, the river health level corresponding to each pixel is determined according to the preset grade division threshold.

[0076] Through calculation and division, not only can the river health condition be objectively quantified to avoid the problem of incomparable results caused by subjective evaluation, but also the spatial distribution analysis of the health condition of a large range of river basins can be realized to provide support for accurately identifying pollution areas.

[0077] In one implementation manner, the weight of each water quality parameter is determined by the objective weighting method, specifically including the following steps: Step S510: Integrate the water quality parameters of the water body area output by the optimal inversion model into a water quality data matrix; Step S520: Standardize each water quality parameter in the water quality data matrix and calculate the weight of each water quality parameter using the entropy weight method.

[0078] In this embodiment, as Figure 2 As shown, in the comprehensive health evaluation stage, the weights need to be calculated using the entropy weight method.

[0079] Based on the global Chl-a, COD, TP, and SDD data matrices obtained through inversion, the objective weights of each parameter are calculated using the entropy weight method.

[0080] The matrix used for entropy weighting calculation is a The matrix, where It is the number of pixels. This refers to the number of evaluation parameters; in this embodiment, the number of evaluation parameters is 4. An example of this matrix is ​​shown in Table 2.

[0081] Table 2

[0082] Subsequently, the weights of each water quality parameter were calculated using the entropy weight method. Specifically, the entropy weight method determines the weights based on the degree of dispersion (variability) of each indicator's data; the more dispersed the data, the smaller the entropy, and the greater the weight.

[0083] It has One sample, Several indicators form a matrix. .

[0084] Perform data standardization, the expression is:

[0085] in, Indicates the first The first sample Individual indicator values, Indicates the first Each indicator in The maximum value among the samples, Indicates the first Each indicator in The minimum value among the samples, These are the standardized matrix elements.

[0086] Calculate the proportion, where It is the first The first indicator The proportion of each sample value is expressed as:

[0087] The entropy value of the first index is calculated The expression is:

[0088] Among them, , to ensure .

[0089] The difference coefficient of the first index is calculated The expression is:

[0090] The smaller the entropy value of the index, the greater the difference coefficient, indicating that the amount of information provided by the index is greater, and a higher weight should be given.

[0091] Finally, the weight is calculated, and the expression is:

[0092] By determining the weight through the entropy weight method, the weight can be allocated based on the dispersion degree of the water quality parameter data itself, avoiding subjective bias in expert experience weighting. For example, when the dispersion degree of total phosphorus is large, the difference coefficient is large, and the corresponding weight is also large, making the final health evaluation result more objective and more comparable.

[0093] In an implementation manner, the calculation of the river health comprehensive index and the river health level of the water body region includes the following steps: Step S530, based on the weight of each water quality parameter and the water quality parameter of the water body region, the river health comprehensive index of each pixel in the water body region is calculated; Step S540, based on the river health comprehensive index and the pre-set health level division threshold, the health level of each pixel in the water body region is divided to obtain the river health level of each pixel in the water body region. Step S550, integrating the river health level of each pixel in the water body region, the river health level spatial distribution map of the research region is obtained.

[0094] In this embodiment, as shown in Figure 2 , finally, the RHHI index and the level division are calculated to realize the generation of the river health level spatial distribution map.

[0095] When calculating the river health comprehensive index and the river health level of the water body region, first, based on the weight of each water quality parameter and the standardized value of the corresponding pixel in the standardized data matrix, the river health comprehensive index (RHHI) of each pixel is calculated through the weighted summation formula.​​

[0096] Specifically, the river health comprehensive evaluation index RHHI value of each pixel is calculated, and the expression is: The index value ranges from 0 to 1, and the larger the value, the better the river health.

[0097] Subsequently, based on the pre-set health level division threshold, the health level of each pixel is divided. Specifically, when the river health comprehensive index is greater than 0.9, it is determined that the river health level of the pixel corresponding region is “excellent”;When is greater than 0.7, it is determined to be “good”;When is greater than 0.5, it is determined to be “mild pollution”;When is greater than 0.3, it is determined to be “moderate pollution”;When is greater than 0.1, it is determined to be “severe pollution”.

[0098] Finally, the river health level of each pixel in the water body region is integrated, and the spatial geographic information of the research region is combined to generate a river health level spatial distribution map of the research region, as shown in Figure 3 . Figure 3 The distribution map shown in the figure identifies different health levels with different colors, such as red for severe pollution areas and green for excellent areas.

[0099] By calculating the river health comprehensive index, multi-dimensional water quality parameters can be converted into a single quantitative index, which is convenient for intuitive measurement of river health. The health level spatial distribution map can clearly show the spatial distribution characteristics of the health condition, helping to quickly identify pollution hotspots and provide accurate decision support for water environment management departments to develop management solutions.

[0100] In summary, the embodiment discloses a river water quality inversion and health evaluation method based on satellite remote sensing. Compared with the prior art, the method has the advantages of high inversion accuracy, objective and scientific evaluation, full-process integration, and realization of macro dynamic monitoring. Specifically, a deep learning model is used to accurately extract water bodies, and a multi-machine learning algorithm is used to optimize the best inversion model, significantly improving the inversion accuracy and robustness. The entropy weight method is introduced to objectively weight according to the discrete degree of water quality parameter data, completely avoiding human subjective bias, making the evaluation results more fair and more comparable. The full-automatic process from remote sensing data input to health level map output is realized, which is efficient and can be operated in business. Make full use of the advantages of remote sensing technology to realize large-scale, long-term dynamic monitoring, accurately identify pollution hotspots, and provide strong decision support for environmental management.

[0101] As Figure 4As shown in the principle block diagram in the foregoing embodiments, the application provides a river water quality inversion and health evaluation system based on satellite remote sensing, which comprises: an original data acquisition module 10, a data preprocessing module 20, a sample set construction module 30, a model selection and verification module 40, and a comprehensive evaluation module 50.

[0102] Specifically, the original data acquisition module 10 is configured to acquire multi-spectral remote sensing images of a research area, and obtain a water quality parameter data set through ground measurement; wherein the multi-spectral remote sensing images and the water quality parameter data set are spatio-temporally matched; the data preprocessing module 20 is configured to preprocess the multi-spectral remote sensing images, and extract images of water body regions in the research area; the sample set construction module 30 is configured to construct a data sample set based on the images of the water body regions and the corresponding water quality parameter data set; the model selection and verification module 40 is configured to construct several inversion models for each water quality parameter in the water quality parameter data set based on the data sample set, and obtain an optimal inversion model corresponding to each water quality parameter through performance verification; and the comprehensive evaluation module 50 is configured to determine the weight of each water quality parameter through an objective weighting method, and calculate a river health comprehensive index and a river health grade of the water body region.

[0103] Based on the foregoing embodiments, the application further provides a terminal device, a principle block diagram of which can be as shown in the principle block diagram in the foregoing embodiments. Figure 5 The terminal device comprises a processor, a memory, a network interface, a display screen, and a temperature sensor connected through a system bus. The processor of the terminal device is configured to provide computing and control capabilities. The memory of the terminal device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The network interface of the terminal device is configured to communicate with external terminals through network connection. The computer program is executed by the processor to implement a river water quality inversion and health evaluation method based on satellite remote sensing. The display screen of the terminal device can be a liquid crystal display screen or an electronic ink display screen. The temperature sensor of the terminal device is pre-set in the terminal device and is configured to detect the running temperature of the internal device.

[0104] Those skilled in the art can understand that, Figure 5 The principle block diagram shown in the foregoing embodiments is only a block diagram of part of the structure related to the application scheme, and does not constitute a limitation on the terminal device to which the application scheme is applied. Specifically, the terminal device can comprise more or fewer components than those shown in the diagram, or combine certain components, or have a different component arrangement.

[0105] In one embodiment, a terminal device is provided that includes memory and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by one or more processors to include instructions for performing the operations of the processes of the embodiments of the above methods.

[0106] A person of ordinary skill in the art can understand that all or part of the processes in the above embodiments can be implemented by using a computer program to instruct related hardware, and the computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, the computer program can include the processes of the embodiments of the above methods. Any reference to memory, storage, database or other medium used in the embodiments of the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct RAM bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0107] Any combination of the technical features of the above embodiments can be made, and in order to make the description concise, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combination of the technical features does not exist, it should be considered as the scope of the present application.

[0108] The above embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent of the present application. It should be pointed out that for a person of ordinary skill in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are all within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A method for river water quality inversion and health evaluation based on satellite remote sensing, characterized in that, The method comprises: acquiring a multispectral remote sensing image of a study area and obtaining a water quality parameter dataset through ground measurement; wherein the multispectral remote sensing image and the water quality parameter dataset are spatio-temporally matched; preprocessing the multispectral remote sensing image and extracting an image of a water body region in the study area; based on the image of the water body region and the corresponding water quality parameter dataset, constructing a data sample set; based on the data sample set, constructing several inversion models for each water quality parameter in the water quality parameter dataset, and obtaining the optimal inversion model corresponding to each water quality parameter through performance verification; determining the weight of each water quality parameter through an objective weighting method and calculating the river health comprehensive index and the river health grade of the water body region.

2. The method according to claim 1, wherein, The acquisition of the multispectral remote sensing image of the study area and the water quality parameter dataset through ground measurement comprises: downloading the multispectral remote sensing image of a specific phase of the study area from a remote sensing satellite data platform; during the period when the multispectral remote sensing image passes through, ground water quality sampling is performed on several sampling points in the study area to obtain the water quality parameters of each sampling point and record the position coordinates of each sampling point; integrating the water quality parameters and corresponding position coordinates of all sampling points to form the water quality parameter dataset; wherein the water quality parameters include at least one of chlorophyll a, chemical oxygen demand, total phosphorus, and transparency.

3. The method according to claim 1, wherein, The preprocessing of the multispectral remote sensing image and the extraction of the image of the water body region in the study area comprise: performing radiation calibration processing and atmospheric correction processing on the multispectral remote sensing image to eliminate the radiation error and atmospheric interference of the image and obtain the surface real reflectance data of the study area; based on the surface real reflectance data, using the normalized water body index to calculate the water body index image of the study area; based on a pre-set water body extraction threshold, performing pixel screening on the water body index image of the study area to retain the pixels corresponding to the water body region and exclude the pixels corresponding to the non-water body region, thereby obtaining the image of the water body region in the study area; wherein the image of the water body region is stored in the form of a water body mask.

4. The method according to claim 1, wherein, The construction of the data sample set based on the image of the water body region and the corresponding water quality parameter dataset comprises: based on the position coordinates of each sampling point in the water quality parameter dataset, locating the pixels corresponding to each sampling point in the image of the water body region and extracting the spectral data of each corresponding pixel; wherein the spectral data includes the reflectance of each band of the multispectral remote sensing image and the spectral index calculated from the reflectance of each band; associating and matching the water quality parameters of each sampling point with the spectral data of the corresponding pixel of the sampling point to form several groups of data pairs of water quality parameters and spectral data; summarizing all data pairs to obtain an initial data set, and randomly dividing the initial data set according to a pre-set proportion to obtain the data sample set composed of a training set and a validation set.

5. The method according to claim 1, wherein, The method comprises the following steps: For each water quality parameter in the water quality parameter dataset, at least two quantitative regression models are constructed based on the training set in the data sample set as candidate inversion models of the water quality parameter; wherein the quantitative regression model includes random forest and support vector regression; The spectral data of each pixel in the validation set in the data sample set is input into each candidate inversion model to calculate the water quality parameter prediction value output by each candidate inversion model; Based on the measured value of the water quality parameter in the validation set in the data sample set and the water quality parameter prediction value output by each candidate inversion model, the determination coefficient, root mean square error and mean absolute error of each candidate inversion model are calculated respectively; The determination coefficient, root mean square error and mean absolute error of each candidate inversion model under the same water quality parameter are compared, and the candidate inversion model with the maximum determination coefficient and the minimum root mean square error and mean absolute error is selected as the optimal inversion model corresponding to the water quality parameter.

6. The satellite remote sensing based river water quality inversion and health evaluation method according to claim 1, characterized in that, The method comprises the following steps: The water quality parameters of the water body region output by the optimal inversion model are integrated into a water quality data matrix; Each water quality parameter in the water quality data matrix is standardized, and the weight of each water quality parameter is calculated by entropy weight method.

7. The method according to claim 6, wherein, The method comprises the following steps: Based on the weight of each water quality parameter and the water quality parameter of the water body region, the river health comprehensive index of each pixel in the water body region is calculated; Based on the river health comprehensive index and the pre-set health grade division threshold, the health grade of each pixel in the water body region is divided to obtain the river health grade of each pixel in the water body region; The river health grades of each pixel in the water body region are integrated to obtain the river health grade spatial distribution map of the research region.

8. A satellite remote sensing-based river water quality inversion and health evaluation system, characterized in that, The system comprises: An original data acquisition module is configured to acquire multispectral remote sensing images of a research region and obtain a water quality parameter dataset through ground measurement; wherein the multispectral remote sensing images and the water quality parameter dataset are spatio-temporally matched; A data preprocessing module is configured to preprocess the multispectral remote sensing images and extract images of a water body region in the research region; A sample set construction module is configured to construct a data sample set based on the images of the water body region and the corresponding water quality parameter dataset; A model selection and verification module is configured to construct several inversion models for each water quality parameter in the water quality parameter dataset based on the data sample set, and obtain the optimal inversion model corresponding to each water quality parameter through performance verification; A comprehensive evaluation module is configured to determine the weight of each water quality parameter by an objective weighting method, and calculate the river health comprehensive index and the river health grade of the water body region.

9. A terminal device, comprising: The terminal device comprises a memory, a processor, and a satellite remote sensing based river water quality inversion and health evaluation program stored in the memory and executable on the processor.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a satellite remote sensing based river water quality inversion and health evaluation program, and the satellite remote sensing based river water quality inversion and health evaluation program is executable on the processor to implement the steps of the satellite remote sensing based river water quality inversion and health evaluation method according to any one of claims 1-7.

Images