Rapid water quality inversion method, device and equipment and storage medium

Through the integration of high-frequency water quality site monitoring data and remote sensing data and the application of machine learning models, the problems of high cost of water quality sampling and insufficient real-time performance are solved, rapid and accurate water quality inversion is achieved, and the efficiency and real-time water quality monitoring are improved. sex.

CN119025927APending Publication Date: 2024-11-26SHANGHAI UBIQUITOUS NAVIGATION TECHNOLOGYCO LTD
View PDF 0 Cites 9 Cited by

Patent Information

Application Number
CN202410762488.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-06-13
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

In the existing water quality remote sensing technology, water quality sampling work is high, long time, and water quality changes rapidly, so the real-time nature of the monitoring results is not enough, which cannot meet the normalized water quality remote sensing monitoring needs.

Method used

By obtaining high-frequency water quality site monitoring data, drone remote sensing data and satellite remote sensing data, performing timing analysis and matching, performing radiation calibration and atmospheric correction processing, building a spectral library data set, and using machine learning models (such as random forests, support Vector machine and XGBoost) perform water quality parameters inversion to achieve rapid water quality inversion.

Benefits of technology

It realizes rapid and precise inversion of water quality parameters, improves the efficiency and real-time nature of remote sensing water quality monitoring, can promptly discover and deal with water quality problems, and provides scientific data support for water environment governance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119025927A_ABST
    Figure CN119025927A_ABST
Patent Text Reader

Abstract

The invention provides a rapid water quality inversion method, apparatus and device, and a storage medium. The method comprises the steps of obtaining high-frequency water quality site monitoring data, unmanned aerial vehicle remote sensing data and high-resolution satellite remote sensing data; carrying out time sequence analysis on the obtained water quality station monitoring data, and carrying out matching based on the obtained unmanned aerial vehicle remote sensing data and satellite remote sensing data to obtain a near-synchronous remote sensing data source; preprocessing the acquired near synchronous remote sensing data source; on the basis of the preprocessed near-synchronous remote sensing data source, stations within the coverage range of the near-synchronous remote sensing data source are obtained, and spectral data in the corresponding near-synchronous remote sensing data source are positioned and extracted on the basis of the position information of the current station; constructing a spectrum database data set based on spectrum data in the near synchronous remote sensing data source, the longitude and latitude of the site, the section name and the water quality parameter; constructing a water quality inversion model, and training the water quality inversion model by using the constructed spectrum library data set; and performing water quality inversion by using the trained water quality inversion model.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of water environment remote sensing technology, specifically, to a fast water quality inversion method, device, equipment and storage medium, and more specifically, to a fast water quality inversion method, device, equipment and storage medium based on online data of monitoring sites. Background Art

[0002] Conventional water quality monitoring methods mainly collect water samples on-site and send them to the laboratory for analysis. The advantage is that the accuracy of water quality monitoring is relatively high, and the disadvantage is that the spatial density is insufficient. Online monitoring methods based on shore or water stations can greatly increase the time and frequency of water quality monitoring, but the construction and maintenance costs are high, and the spatial distribution is more limited. The development of remote sensing technology has provided a new idea for water quality monitoring. By establishing the relationship between the water body reflectance spectral characteristics and the water quality parameter concentration, an inversion model is constructed to form a point-to-surface monitoring system with the advantages of fast response, low cost, and real-time dynamic monitoring. It can realize the monitoring of a large range of waters and meet the needs of routine monitoring and emergency monitoring. Compared with manual sampling and analysis, it can make basic judgments and trend analysis on the water quality status of the entire river in a shorter time.

[0003] Remote sensing image inversion of water quality is to use the response of water body optical characteristics to water quality parameters to achieve the purpose of obtaining water quality distribution. The current inversion methods mainly include analytical methods, empirical methods and semi-empirical methods. At present, analytical methods, empirical methods and semi-empirical methods in remote sensing water quality inversion need to combine various types of water quality spectral information as the basic data for model construction. Water quality spectral information is indispensable. Therefore, each time remote sensing water quality inversion is carried out, a lot of manpower, material resources and financial resources are spent on water quality sampling work that is nearly synchronized with satellite or drone transit shooting. Now, water quality inspection and testing consumes a long time, water quality changes rapidly, the real-time nature of the monitoring results is not enough, and the field sampling data produced has few opportunities for later reuse, which is not suitable for normalized water quality remote sensing monitoring. At the same time, the data from water quality monitoring stations can well reflect the problems of water quality points in rivers and lakes, but cannot well reflect the problems on the macro level of river sections. Summary of the invention

[0004] In view of the defects in the prior art, the object of the present invention is to provide a rapid water quality inversion method, device, equipment and storage medium.

[0005] A first aspect of the present invention provides a rapid water quality inversion method, comprising:

[0006] Access high-frequency water quality site monitoring data, drone remote sensing data, and high-resolution satellite remote sensing data;

[0007] Conduct time series analysis on the acquired water quality site monitoring data, and match and acquire near-synchronous remote sensing data sources based on the image data time of the acquired UAV remote sensing data and satellite remote sensing data;

[0008] Preprocessing the acquired near-synchronous remote sensing data source to obtain a preprocessed near-synchronous remote sensing data source;

[0009] Based on the preprocessed near-synchronous remote sensing data source, the stations within the coverage of the near-synchronous remote sensing data source are obtained, and the spectral data in the corresponding near-synchronous remote sensing data source are located and extracted based on the location information of the current station;

[0010] A spectral library dataset is constructed based on the spectral data, longitude and latitude of the station, section name and water quality parameters in the near-synchronous remote sensing data source;

[0011] A water quality inversion model is constructed, and the water quality inversion model is trained using the constructed spectral library data set to obtain a trained water quality inversion model;

[0012] Using the trained water quality inversion model to perform water quality inversion;

[0013] Among them, the water quality online monitoring site data includes: permanganate index, ammonia nitrogen, total phosphorus water quality indicators, pH value, dissolved oxygen, water temperature, turbidity, conductivity parameters, instantaneous flow velocity, instantaneous flow and cross-sectional area data.

[0014] Optionally, in a first implementation of the first aspect of the present invention, the time series analysis of the acquired water quality site monitoring data is performed, and based on the image data time of the acquired drone remote sensing data and the satellite remote sensing data, a near-synchronous remote sensing data source is matched and acquired, including:

[0015] Perform time series analysis on the acquired water quality site monitoring data; obtain the water quality site monitoring data collection time closest to the image data time, and obtain the corresponding water quality site monitoring data based on the closest water quality site monitoring data collection time; build a near-synchronous remote sensing data source based on the UAV remote sensing data and satellite remote sensing data corresponding to the current image data time.

[0016] Optionally, in a second implementation of the first aspect of the present invention, preprocessing the acquired near-synchronous remote sensing data source to obtain a preprocessed near-synchronous remote sensing data source includes:

[0017] The acquired near-synchronous remote sensing data source is subjected to radiation calibration and atmospheric correction processing to obtain a pre-processed near-synchronous remote sensing data source;

[0018] The radiation calibration is to convert the original DN value of remote sensing impact into apparent reflectance or absolute radiation brightness, so as to achieve standardization and accuracy processing of image data;

[0019] The atmospheric correction is that the 6S model simulates the surface reflectivity based on geometric parameters, atmospheric mode, aerosol mode, optical thickness, target altitude and sensor altitude, and compares the simulated surface reflectivity with the actual remote sensing data to perform atmospheric correction.

[0020] Optionally, in a third implementation of the first aspect of the present invention, the method of acquiring a site within the coverage of the near-synchronous remote sensing data source based on the preprocessed near-synchronous remote sensing data source, and locating and extracting spectral data in the corresponding near-synchronous remote sensing data source based on the location information of the current site includes:

[0021] According to the four-axis coordinate range information in the near-synchronous remote sensing data source, the stations within the coverage range of the near-synchronous remote sensing data source are judged; according to the rational polynomial coefficients provided in the remote sensing image data, the forward projection expression in the RPC model is used to convert the longitude and latitude coordinates of the stations within the coverage range into row and column coordinates on the remote sensing image data, and the spectral data is extracted based on the row and column coordinates on the remote sensing image.

[0022] Optionally, in a fourth implementation of the first aspect of the present invention, the constructing of a quantitative inversion model for water quality parameters includes: constructing a water quality inversion model based on key characteristic variables that meet preset requirements;

[0023] The water quality inversion model is based on machine learning including random forest regression, support vector machine regression or XGBoost regression to achieve water quality parameter estimation of water bodies;

[0024] The key characteristic variables that meet the preset requirements include:

[0025] Extract the single-band reflectance data of remote sensing images corresponding to each sampling point, select any two bands for band combination, and obtain the potential characteristic variables related to the concentration of water quality parameters;

[0026] The potential characteristic variables are analyzed for correlation to obtain the key characteristic variables;

[0027] The potential characteristic variables are subjected to correlation analysis to obtain key characteristic variables, including: performing Pearson correlation analysis on the measured data of each water quality parameter and band combination to determine the key characteristic variables for constructing a water quality parameter inversion model.

[0028] Optionally, in a fifth implementation of the first aspect of the present invention, the method further comprises: implementing water quality classification based on a spectral angle matching method;

[0029] A spectral library of known water quality categories was constructed using different remote sensing data sources and water quality site monitoring data;

[0030] The current water quality category spectral library is used as the reference spectral library;

[0031] Compare the spectral vector of each pixel in the remote sensing image with the reference spectral vector, and measure the similarity between the spectra by calculating the angle between the two spectral vectors;

[0032] The angle between the two spectral vectors is:

[0033]

[0034] Among them, θ x,y represents the angle between the reference spectrum vector X and the spectrum vector Y to be measured; n represents the number of bands of the spectrum curve; Xi represents the spectral reflectance value of the reference spectrum vector in the i-th band; Yi represents the spectral reflectance value of the spectrum vector to be measured in the i-th band.

[0035] Optionally, in a sixth implementation of the first aspect of the present invention, the method further includes: using a determination coefficient R 2 The water quality inversion model was evaluated by the root mean square error (RMSE);

[0036] Using the determination coefficient R 2 Measure the fit of the water quality inversion model;

[0037] The root mean square error RMSE is used to measure the deviation between the estimated value and the actual value;

[0038]

[0039] Among them, y' j represents the predicted value; represents the average value of the measured values; j represents the measured value; n represents the sample size;

[0040] Based on the confusion matrix method, the water quality category was evaluated by using the water sample images of the verification points through the overall classification accuracy, recall rate and precision rate.

[0041] The overall classification accuracy includes:

[0042]

[0043] The recall rate includes:

[0044]

[0045] The accuracy includes:

[0046]

[0047] Among them, TP represents a positive sample predicted to be positive; TN represents a negative sample predicted to be negative; FP represents a negative sample predicted to be positive; and FN represents a positive sample predicted to be negative.

[0048] A second aspect of the present invention provides a rapid water quality inversion device, comprising:

[0049] Data acquisition module, used to acquire high-frequency water quality site monitoring data, drone remote sensing data, and high-resolution satellite remote sensing data;

[0050] The near-synchronous remote sensing data source acquisition module is used to perform time series analysis on the acquired water quality site monitoring data, and match and acquire the near-synchronous remote sensing data source based on the image data time of the acquired UAV remote sensing data and satellite remote sensing data;

[0051] A preprocessing module is used to preprocess the acquired near-synchronous remote sensing data source to obtain a preprocessed near-synchronous remote sensing data source;

[0052] The spectral data acquisition module is used to acquire the sites within the coverage of the near-synchronous remote sensing data source based on the pre-processed near-synchronous remote sensing data source, and locate and extract the spectral data in the corresponding near-synchronous remote sensing data source based on the location information of the current site;

[0053] A spectral library dataset construction module is used to construct a spectral library dataset based on the spectral data in the near-synchronous remote sensing data source, the longitude and latitude of the station, the section name, and the water quality parameters;

[0054] A water quality inversion model training module is used to construct a water quality inversion model and train the water quality inversion model using the constructed spectral library data set to obtain a trained water quality inversion model;

[0055] A water quality inversion module is used to perform water quality inversion using the trained water quality inversion model;

[0056] Among them, the water quality online monitoring site data includes: permanganate index, ammonia nitrogen, total phosphorus water quality indicators, pH value, dissolved oxygen, water temperature, turbidity, conductivity parameters, instantaneous flow velocity, instantaneous flow and cross-sectional area data.

[0057] Optionally, in a first implementation of the second aspect of the present invention, the near-synchronous remote sensing data source acquisition module includes:

[0058] Perform time series analysis on the acquired water quality site monitoring data; obtain the water quality site monitoring data collection time closest to the image data time, and obtain the corresponding water quality site monitoring data based on the closest water quality site monitoring data collection time; build a near-synchronous remote sensing data source based on the UAV remote sensing data and satellite remote sensing data corresponding to the current image data time.

[0059] Optionally, in a second implementation of the second aspect of the present invention, the preprocessing module includes:

[0060] The acquired near-synchronous remote sensing data source is subjected to radiation calibration and atmospheric correction processing to obtain a pre-processed near-synchronous remote sensing data source;

[0061] The radiation calibration is to convert the original DN value of remote sensing impact into apparent reflectance or absolute radiation brightness, so as to achieve standardization and accuracy processing of image data;

[0062] The atmospheric correction is that the 6S model simulates the surface reflectivity based on geometric parameters, atmospheric mode, aerosol mode, optical thickness, target altitude and sensor altitude, and compares the simulated surface reflectivity with the actual remote sensing data to perform atmospheric correction.

[0063] Optionally, in a third implementation of the second aspect of the present invention, the spectral data acquisition module includes:

[0064] According to the four-axis coordinate range information in the near-synchronous remote sensing data source, the stations within the coverage range of the near-synchronous remote sensing data source are judged; according to the rational polynomial coefficients provided in the remote sensing image data, the forward projection expression in the RPC model is used to convert the longitude and latitude coordinates of the stations within the coverage range into row and column coordinates on the remote sensing image data, and the spectral data is extracted based on the row and column coordinates on the remote sensing image.

[0065] Optionally, in a fourth implementation of the second aspect of the present invention, the constructing of a quantitative inversion model for water quality parameters includes: constructing a water quality inversion model based on key characteristic variables that meet preset requirements;

[0066] The water quality inversion model is based on machine learning including random forest regression, support vector machine regression or XGBoost regression to achieve water quality parameter estimation of water bodies;

[0067] The key characteristic variables that meet the preset requirements include:

[0068] Extract the single-band reflectance data of remote sensing images corresponding to each sampling point, select any two bands for band combination, and obtain the potential characteristic variables related to the concentration of water quality parameters;

[0069] The potential characteristic variables are analyzed for correlation to obtain the key characteristic variables;

[0070] The potential characteristic variables are subjected to correlation analysis to obtain key characteristic variables, including: performing Pearson correlation analysis on the measured data of each water quality parameter and band combination to determine the key characteristic variables for constructing a water quality parameter inversion model.

[0071] Optionally, in a fifth implementation of the second aspect of the present invention, the device further comprises: a water quality classification module, configured to classify water quality based on a spectral angle matching method;

[0072] A spectral library of known water quality categories was constructed using different remote sensing data sources and water quality site monitoring data;

[0073] The current water quality category spectral library is used as the reference spectral library;

[0074] Compare the spectral vector of each pixel in the remote sensing image with the reference spectral vector, and measure the similarity between the spectra by calculating the angle between the two spectral vectors;

[0075] The angle between the two spectral vectors is:

[0076]

[0077] Among them, θ x,y represents the angle between the reference spectrum vector X and the spectrum vector Y to be measured; n represents the number of bands of the spectrum curve; Xi represents the spectral reflectance value of the reference spectrum vector in the i-th band; Yi represents the spectral reflectance value of the spectrum vector to be measured in the i-th band.

[0078] Optionally, in a sixth implementation of the second aspect of the present invention, the device further comprises: a model evaluation module for using the determination coefficient R 2 The water quality inversion model was evaluated by the root mean square error (RMSE);

[0079] Using the determination coefficient R 2 Measure the fit of the water quality inversion model;

[0080] The root mean square error RMSE is used to measure the deviation between the estimated value and the actual value;

[0081]

[0082] Among them, y' j represents the predicted value; represents the average value of the measured values; j represents the measured value; n represents the sample size;

[0083] The water quality category evaluation module is used to evaluate the water quality category using the water sample images of the verification points based on the confusion matrix method through the overall classification accuracy, recall rate and precision;

[0084] The overall classification accuracy includes:

[0085]

[0086] The recall rate includes:

[0087]

[0088] The accuracy includes:

[0089]

[0090] Among them, TP represents a positive sample predicted to be positive; TN represents a negative sample predicted to be negative; FP represents a negative sample predicted to be positive; and FN represents a positive sample predicted to be negative.

[0091] A third aspect of the present invention is an electronic device, comprising a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory to enable the electronic device to execute each step of the above-mentioned rapid water quality inversion method.

[0092] A fourth aspect of the present invention is a computer-readable storage medium having instructions stored thereon, wherein the instructions, when executed by a processor, implement the various steps of the above-mentioned rapid water quality inversion method.

[0093] Compared with the prior art, the present invention has the following beneficial effects:

[0094] 1. The present invention deeply integrates the water quality site monitoring data, UAV remote sensing data and satellite remote sensing data acquired at high frequency, and matches the online water quality monitoring data that is closest in time to the image data through precise time series analysis and matching, so as to realize the fusion and complementarity of multi-dimensional data; fully utilizes the longitude and latitude information provided by the water quality site monitoring data, accurately locates and extracts the corresponding spectral information of the remote sensing data source; gradually establishes a water quality characteristic spectral library from different remote sensing data sources by continuously accumulating real-time online monitoring data; this spectral library not only contains rich water quality information, but also provides strong data support for the quantitative inversion of water quality parameters, and provides a more scientific and accurate basis for water quality monitoring and water environment management;

[0095] 2. Based on different data source spectral libraries and different machine learning water quality parameter inversion models, the present invention obtains the accuracy evaluation results of different machine learning inversion models for different water quality parameters through spectral index construction and characteristic variable screening, inversion model construction and model verification; for the same water quality parameter, the determination coefficient R is selected. 2 The highest machine learning model is used as the optimal inversion model for the water quality parameter; based on the quantitative inversion model of the water quality parameter with the highest prediction accuracy and the water index reflectivity map of the remote sensing image, a multi-parameter water quality inversion with full coverage of regional water quality is achieved;

[0096] 3. The present invention accumulates a water body spectral library based on water quality station data to form a better model training sample. After each acquisition of remote sensing image data, it can achieve efficient model construction and regional water quality inversion, realize "inversion with image", improve the efficiency of remote sensing water quality inversion, and provide scientific data support for water ecological environment monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0097] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0098] Figure 1 A first flow chart of a rapid water quality inversion method provided in an embodiment of the present invention.

[0099] Figure 2 A second flow chart of the rapid water quality inversion method provided in an embodiment of the present invention.

[0100] Figure 3 The present invention is a flowchart of preprocessing the acquired near-synchronous remote sensing data source in an embodiment of the present invention.

[0101] Figure 4 Schematic diagram of the spectral vector angle in an embodiment of the present invention.

[0102] Figure 5 A schematic structural diagram of a rapid water quality inversion device provided in an embodiment of the present invention.

[0103] Figure 6 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0104] The embodiment of the present invention provides a method, device, equipment and storage medium for rapid water quality inversion, including: obtaining water quality site monitoring data, unmanned aerial vehicle remote sensing data and satellite remote sensing data based on a high frequency that meets preset requirements; performing time series analysis on the acquired water quality site monitoring data, and matching and acquiring near-synchronous remote sensing data sources based on the acquired unmanned aerial vehicle remote sensing data and satellite remote sensing data; preprocessing the acquired near-synchronous remote sensing data sources to obtain near-synchronous remote sensing data sources after preprocessing; obtaining sites within the coverage of the near-synchronous remote sensing data sources based on the preprocessed near-synchronous remote sensing data sources, locating and extracting spectral data in the corresponding near-synchronous remote sensing data sources based on the location information of the current sites; constructing a spectral library data set based on the spectral data in the near-synchronous remote sensing data sources, the longitude and latitude of the sites, the section name and the water quality parameters; constructing a water quality inversion model, and using the constructed spectral library data set to train the water quality inversion model to obtain the trained water quality inversion model; and using the trained water quality inversion model to perform water quality inversion. The present invention solves the problems in the prior art that the water quality sampling work is costly and time-consuming, and the water quality changes rapidly, resulting in insufficient real-time monitoring results.

[0105] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0106] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 The first embodiment of the rapid water quality inversion method in the embodiment of the present invention includes:

[0107] 101. Obtain water quality site monitoring data, drone remote sensing data, and satellite remote sensing data at a high frequency that meets preset requirements;

[0108] In this embodiment, the data resources of water quality monitoring stations are fully utilized to quickly respond to point problems in river and lake water quality. In order to ensure the timeliness and accuracy of the data, high-frequency data acquisition is performed, and water quality section monitoring station data is collected every 4 hours. High-frequency data updates can grasp the water quality status in real time, provide strong support for the timely discovery and treatment of water quality problems, and provide data support for subsequent multi-source data matching.

[0109] The data from water quality monitoring stations include key water quality indicators such as permanganate index, ammonia nitrogen, and total phosphorus. These indicators can intuitively reflect the degree of pollution and nutritional status of the water body. At the same time, parameters such as pH value, dissolved oxygen, water temperature, turbidity, and conductivity are obtained, which helps to fully understand the chemical and physical properties of the water body. In addition, the instantaneous flow velocity, instantaneous flow rate, and cross-sectional area data obtained help to more deeply understand the flow characteristics and hydrodynamic conditions of the water body. Through the data acquisition and monitoring method proposed in this patent, it is possible to more accurately grasp the changes in the water quality of rivers and lakes, and to promptly discover and deal with water quality problems from point to surface.

[0110] 102. Perform time series analysis on the acquired water quality site monitoring data, and match and acquire near-synchronous remote sensing data sources based on the acquired UAV remote sensing data and satellite remote sensing data;

[0111] In this embodiment, the continuity and real-time characteristics of the water quality station monitoring data are fully utilized, and the near-synchronous remote sensing data source is accurately matched and obtained by performing time series analysis on the water quality station monitoring data. In order to achieve near-synchronous remote sensing data source acquisition, the matching principle is to ensure that the acquisition time of the remote sensing image is as close as possible to the acquisition time of the water quality station data.

[0112] In this embodiment, the spatial resolution and spectral characteristics of the remote sensing image are considered simultaneously, and appropriate remote sensing satellites and sensors are selected to obtain remote sensing images with high resolution and high spectral information to ensure their matching and reliability with the water quality site data.

[0113] 103. Preprocessing the acquired near-synchronous remote sensing data source to obtain a preprocessed near-synchronous remote sensing data source;

[0114] In this embodiment, the remote sensing images are subjected to radiometric calibration and atmospheric correction batch preprocessing. Among them, radiometric calibration is one of the key steps in remote sensing image preprocessing, which aims to eliminate the performance differences of the sensor itself and the influence of the external environment on the image data, ensure the accuracy and reliability of the data, and enable it to truly reflect the reflective characteristics of the ground objects. Among them, atmospheric correction is another indispensable preprocessing step. Due to the presence of various factors in the atmosphere (such as scattering, absorption, etc.), remote sensing images will be subject to varying degrees of interference during transmission. In order to eliminate these interferences, an atmospheric correction algorithm based on the 6S model is used to batch process the images to restore the true reflectivity of the ground objects.

[0115] 104. Based on the preprocessed near-synchronous remote sensing data source, obtain a site within the coverage of the near-synchronous remote sensing data source, locate and extract spectral data in the corresponding near-synchronous remote sensing data source based on the location information of the current site;

[0116] In this embodiment, the water quality stations located within the image coverage are determined based on the four-axis coordinate range information of the image, and the longitude and latitude coordinates of the station data within the range are converted into image row and column coordinates according to the RPC model to extract the spectral band information.

[0117] 105. Construct a spectral library dataset based on the spectral data, longitude and latitude of the station, section name and water quality parameters in the near-synchronous remote sensing data source;

[0118] 106. Constructing a water quality inversion model, and using the constructed spectral library data set to train the water quality inversion model to obtain a trained water quality inversion model;

[0119] In this embodiment, the water quality inversion model is constructed based on key characteristic variables that meet preset requirements;

[0120] The water quality inversion model is based on machine learning including random forest regression, support vector machine regression or XGBoost regression to achieve water quality parameter estimation of water bodies;

[0121] The key characteristic variables that meet the preset requirements include:

[0122] Extract the single-band reflectance data of remote sensing images corresponding to each sampling point, select any two bands for band combination, and obtain the potential characteristic variables related to the concentration of water quality parameters;

[0123] The potential characteristic variables are analyzed for correlation to obtain the key characteristic variables;

[0124] The potential characteristic variables are subjected to correlation analysis to obtain key characteristic variables, including: performing Pearson correlation analysis on the measured data of each water quality parameter and band combination to determine the key characteristic variables for constructing a water quality parameter inversion model.

[0125] 107. Use the trained water quality inversion model to perform water quality inversion.

[0126] See also Figure 2 The second embodiment of the rapid water quality inversion method of the express in the embodiment of the present invention includes:

[0127] 201. Acquire water quality site monitoring data, drone remote sensing data, and satellite remote sensing data at a high frequency that meets preset requirements;

[0128] 202. Perform time series analysis on the acquired water quality site monitoring data, and match and acquire near-synchronous remote sensing data sources based on the acquired UAV remote sensing data and satellite remote sensing data;

[0129] In this embodiment, the water quality site monitoring data, UAV remote sensing data and satellite remote sensing data acquired at a high frequency are deeply integrated, and through precise time series analysis and matching, the water quality online monitoring data that is closest in time to the image data is matched to achieve the fusion and complementarity of multi-dimensional data.

[0130] 203. Preprocessing the acquired near-synchronous remote sensing data source to obtain a preprocessed near-synchronous remote sensing data source;

[0131] In this embodiment, if Figure 3 As shown, the acquired near-synchronous remote sensing data source is subjected to radiation calibration and atmospheric correction processing to obtain a pre-processed near-synchronous remote sensing data source;

[0132] Radiometric calibration is when the user needs to calculate the spectral reflectance or spectral radiance of the object, or needs to compare images acquired at different times and with different sensors. The brightness grayscale value of the image must be converted into absolute radiance. This process is radiometric calibration, which is divided into absolute calibration and relative calibration. Absolute calibration refers to the use of various standard radiation sources to establish a quantitative relationship between the spectral radiance value at the entrance pupil of the imaging spectrometer and the digital quantization value output by the imaging spectrometer in different spectral bands. Relative calibration refers to determining the relative value of the radiation measured between each pixel in the scene, between each detector, between each spectrum, and at different times. The principle of radiometric calibration converts the original DN value of the remote sensing image into apparent reflectance; and converts the brightness grayscale value of the image into absolute radiance.

[0133] L λ =Gain*DN+Offset

[0134] Where, L λ is the radiation brightness, Gain is the radiation gain, and Offset is the radiation bias.

[0135] Atmospheric correction is a 6S model that simulates the surface reflectivity based on geometric parameters, atmospheric model, aerosol model, optical thickness, target altitude and sensor altitude, and compares the simulated surface reflectivity with the actual remote sensing data to perform atmospheric correction.

[0136] The geometric parameters include four variables: solar zenith angle, solar azimuth angle, observation zenith angle, and observation azimuth angle, which are obtained through RPC files.

[0137] The atmospheric mode includes: selecting a corresponding atmospheric mode according to the actual situation of the image, as shown in Table 1.

[0138] Table 1

[0139]

[0140]

[0141] Aerosol type parameters: Select the corresponding aerosol type according to the actual situation of the image, as shown in Table 2.

[0142] Table 2

[0143]

[0144] Target height parameter: The default is sea level height. This parameter is set to the average target height of the area calculated from the DEM data product of the image area.

[0145] Sensor altitude parameters, spectral parameters: determined by satellite type.

[0146] 204. Based on the preprocessed near-synchronous remote sensing data source, obtain a site within the coverage of the near-synchronous remote sensing data source, locate and extract spectral data in the corresponding near-synchronous remote sensing data source based on the location information of the current site;

[0147] In this embodiment, the construction of the spectral library requires spectral data on the remote sensing image. In order to extract the remote sensing reflectance of each band in the pre-processed data, the longitude and latitude coordinates of the water quality monitoring site need to be converted into row and column coordinates on the remote sensing image. According to the rational polynomial coefficients (RPC) provided in the remote sensing image data, the longitude and latitude are converted into row and column numbers using the forward projection expression in the RPC model to obtain the spectral data of each band corresponding to the site and the corresponding water quality parameter monitoring data.

[0148] The RPC model is a high-precision fit of a strict geometric imaging model in a mathematical sense. It associates the geodetic coordinates D (Latitude, Longitude, Height) of a ground point with its corresponding image point coordinates d (Line, Sample) using a ratio polynomial. For a remote sensing image, the following ratio polynomial is defined:

[0149]

[0150] Where: rn, cn are the normalized image row and column coordinates, P, L, H are the normalized ground longitude, latitude and elevation, respectively, where:

[0151]

[0152] Substitute the offset coefficient and scaling coefficient in the remote sensing image RPC header file to obtain the coordinate conversion relationship between the image row and column coordinates and the latitude and longitude coordinates. The formula is as follows:

[0153] r=r n × s +r0

[0154] c=c n ×c s +c0

[0155] Where: r s With c s is the scaling factor, r0 and c0 represent the offset coefficients, and r and c are the row and column numbers of the image.

[0156] 205. Construct a spectral library dataset based on the spectral data, longitude and latitude of the station, section name and water quality parameters in the near-synchronous remote sensing data source;

[0157] In this embodiment, sensors carried by different sensors and platforms can provide data with different spectral ranges and spectral response characteristics, thereby obtaining more diverse spectral band information of water bodies. At the same time, data collected by different platforms have different band advantages and limitations. By accumulating spectral libraries from different data sources, the quality and reliability of the overall data can be improved, and the generalization and robustness of the water quality inversion model can be enhanced. In addition, a single platform is often difficult to take into account both high spatial resolution and high temporal resolution. By integrating multi-source remote sensing platform data, higher resolution can be achieved in space and time, thereby more finely monitoring and analyzing the dynamic changes of water quality.

[0158] Satellite remote sensing scale

[0159] The satellite model, sensor type and number, and imaging date and time are used as the key fields of the satellite remote sensing scale. According to the four-coordinate range in the image header file information, it is judged whether there is a water quality monitoring station. The spectral information of each band on the image where the station is located, as well as the longitude and latitude of the corresponding station, section name, water quality parameters and other data are obtained. Finally, a spectral library dataset at the satellite remote sensing scale is constructed.

[0160] UAV remote sensing scale

[0161] In the UAV remote sensing scale, the UAV is divided into three sensor types: visible light, multispectral, and hyperspectral according to the lens parameters carried by the UAV. Based on the sensor, the band spectral information on the UAV image where the station is located and the corresponding station data are determined and obtained, and finally the spectral library dataset in the UAV remote sensing scale is constructed.

[0162] 206. Constructing a water quality inversion model, and using the constructed spectral library data set to train the water quality inversion model to obtain a trained water quality inversion model;

[0163] In this embodiment, the water quality inversion model is constructed based on key characteristic variables;

[0164] In this embodiment, spectral index construction and characteristic variable screening include:

[0165] The typical spectral index in water quality remote sensing inversion is a two-band combination. The band combination is formed by band ratio, addition, subtraction and normalized index calculation, which can reduce some noise interference, enhance the sensitivity of spectral reflectance to water quality parameters, and greatly improve the prediction accuracy of water quality inversion model. Therefore, this paper first extracts the single-band reflectance data of remote sensing images corresponding to each sampling point, selects any two bands to perform the following band combination calculations: Bi-Bj, Bi+Bj, Bi / Bj and (Bi-Bj) / (Bi+Bj) (i and j values ​​can be integers, i≠j), and obtains the potential characteristic variables related to the concentration of water quality parameters. The characteristic variable combination and spectral index calculation formula are shown in Table 3.

[0166] Table 3

[0167] Combination Indicator name Calculation formula Single band <![CDATA[B i ]]> <![CDATA[B1,B2,B3……B i ]]> Sum band combination <![CDATA[SSI (i,j) ]]> <![CDATA[SSI (i,j) =B i +B j ]]> Difference band combination <![CDATA[DSI (i,j) ]]> <![CDATA[DSI (i,j) =B i -B j ]]> Normalized band combination <![CDATA[NSI (i,j) ]]> <![CDATA[NSI (i,j) =(B i -B j ) / (B i +B j )]]> Ratio Band Combination <![CDATA[RSI (i,j) ]]> <![CDATA[RSI (i,j) =B i / B j ]]>

[0168] In order to improve the performance of the model, a correlation analysis is performed before modeling to determine the key characteristic variables of the model. The statistical linear correlation between two variables can be determined using the Pearson correlation coefficient method. In this embodiment, Pearson correlation analysis is performed on the measured data of each water quality parameter and band combination to determine the key characteristic variables for constructing the water quality parameter inversion model. In general, selecting a band or band combination with higher correlation as the model input can improve the fitting ability and accuracy of the model. Therefore, the five spectral indices with the highest correlation with the water quality parameters can be selected as the key characteristic variables for the machine learning modeling input.

[0169] Construction of water quality parameter inversion models based on different machine learning

[0170] Random Forest Regression

[0171] Random forest (RF) is an ensemble learning algorithm for classification and regression based on classification trees. The random forest regression (RFR) model randomly samples the sample observations and feature variables of the modeling data set. In the process of generating many decision trees, each sampling result is a tree, and each tree will generate rules and judgment values ​​that conform to its own attributes. The forest will eventually integrate the rules and judgment values ​​of all decision trees together to achieve regression of the random forest algorithm. RFR represents the complex relationship between the response variable and a set of explanatory variables. It has the advantages of reducing overfitting robustness, low sensitivity to parameter adjustment, fewer parameters that need to be adjusted, and strong noise resistance. The steps to build a random forest based on the bagging framework using the RandomForest function in R software for RF regression model are as follows:

[0172] 1) First, divide the data into training set and validation set;

[0173] 2) Randomly select training samples from the training set and back-test them, repeat N times, to obtain N sub-training sets, which are used to build N decision trees;

[0174] 3) Randomly select m(m≤M try ) specific variables to determine the best split point, M try is the number of input independent variables;

[0175] 4) Use the sub-training set to train N decision trees, use the validation set for regression prediction, and obtain N results;

[0176] 5) Take the average of N results as the final result of random forest.

[0177] The two key parameters in the random forest modeling process are the number of feature variables (M) and try ) and the number of decision trees (N tree ), which greatly affects the accuracy of modeling. In order to optimize the random forest model, the Caret package in R language was used to tune the parameters, and the grid search method was used to determine the optimal parameters.

[0178] Support Vector Machine Regression

[0179] Support vector regression (SVR) uses nonlinear mapping to map data into a high-dimensional data feature space, so that the independent variables and dependent variables in the high-dimensional data feature space have good linear regression characteristics, and then return to the original space after fitting in the feature space.

[26] SVR has good robustness and is tolerant to noise and outliers in the data. The basic principles of support vector machine regression can be summarized into the following steps:

[0180] 1) Data preprocessing: First, preprocess the data, including missing value processing and normalization.

[0181] 2) Select kernel function: Commonly used kernel functions include linear kernel, radial basis function kernel and polynomial kernel.

[0182] 3) Model training: Use the training dataset (X, Y) to train the support vector machine regression model.

[0183] 4) Model evaluation: Use the validation dataset (X v , Y v )Evaluate the performance of the model, such as the difference between the predicted value and the true value.

[0184] 5) Parameter adjustment: According to the model evaluation results, adjust the model parameters, such as the regularization parameter C, kernel function parameters, etc., to optimize the model performance.

[0185] 6) Model application: Use the trained support vector machine regression model to perform regression prediction on new data.

[0186] The support vector machine model was trained by calling the svm function in the e1071 package through R software, and the radial basis kernel function was used for support vector regression. The important parameters C (penalty coefficient) and gamma (kernel function coefficient) of the radial basis support vector machine model were adjusted by the grid search method.

[0187] XGBoost Regression

[0188] The XGBoost algorithm is an efficient gradient boosting decision tree algorithm. The core of the algorithm is to use multiple weak learners to build a strong learner by gradually optimizing the loss function. Specifically, each weak learner is a decision tree model, and XGBoost uses a custom loss function to consider both the size and complexity of the error in the process of building each tree. Compared with the traditional gradient boosting algorithm, XGBoost improves and optimizes the model by adding regularization terms, thereby improving operating efficiency and preventing overfitting of the model, while supporting research on a small number of samples.

[0189] The process of XGBoost model building and prediction is implemented by directly calling the xgboost function in the xgboost package of R language. The basic process is: 1) import data and divide the data into two parts: training set and test set; 2) perform model parameter tuning; 3) train the model: after determining the optimal parameter configuration of the model, use the xgboost function to build the training model; 4) use the predict function to perform prediction.

[0190] There are several important parameters in the XGBoost regression process: learning rate eta, step size of iterative decision tree, and maximum depth of the tree. The smaller the step size, the slower the training speed and the easier it is for the model to overfit. As the maximum depth value increases, the recorded model will become more complex. The value of gamma ranges from 0 to 1. The larger the gamma, the more conservative the algorithm. During the modeling process, the grid search method is used to tune the parameters that have the greatest impact on the model. The maximum depth is set to 3 to specify the maximum depth of the tree, while the learning rate eta is set to 0.3 to prevent overfitting. The gamma value is set to 0, which controls the regularization term of the model complexity weight value. Subsample is set to 1 to extract 100% of the data for training tree growth.

[0191] 207. Perform water quality inversion using the trained water quality inversion model;

[0192] 208. Water quality classification based on spectral angle matching method;

[0193] A spectral library of known water quality categories constructed using data from different remote sensing data sources and water quality monitoring stations is used as a reference spectral library. The spectral vector of each pixel in the remote sensing image is compared with the reference spectral vector, and the similarity between the spectra is measured by calculating the angle between the two spectral vectors. The smaller the angle, the higher the similarity.

[0194] like Figure 4 As shown, the formula for calculating the cosine angle between two spectral vectors is as follows:

[0195]

[0196] Among them, θ x,y represents the angle between the reference spectrum vector X and the spectrum vector Y to be measured; n represents the number of bands of the spectrum curve; Xi represents the spectral reflectance value of the reference spectrum vector in the i-th band; Yi represents the spectral reflectance value of the spectrum vector to be measured in the i-th band.

[0197] The angle θ between all image pixels and each set of spectral curves in the reference spectral library is obtained by spectral angle calculation, and the minimum angle θ between each image pixel and all spectral curves in the reference spectral library is selected. min The corresponding spectral curve takes the water quality category corresponding to the selected reference spectral curve as the water quality classification result of the image pixel, thereby realizing rapid classification of water quality of unknown remote sensing images.

[0198] 209. Using the coefficient of determination R 2 The water quality inversion model was evaluated by the root mean square error (RMSE);

[0199] 70% of the water sample data sets from randomly selected water quality monitoring stations were divided into training sets, and the remaining 30% of the water sample data sets were used as test sets. The five key characteristic variables with the strongest correlation with water quality parameters determined in the correlation analysis were used as the input of the model, and the concentration values ​​of water quality parameters were used as the output variables of the model. The machine learning models of different water quality parameters were constructed and the accuracy was evaluated. The predicted values ​​of water quality parameters obtained based on the machine learning model were regressed and fitted with the test values, and the prediction performance of different models was compared according to the model accuracy evaluation index.

[0200] In order to obtain the optimal inversion model for each water quality parameter, the coefficient of determination (R 2 ) and root mean squared error (RMSE) were used to evaluate the model accuracy of each machine learning algorithm model for water quality inversion. 2 Represents the degree of fit of the model, ranging from 0 to 1, R 2The closer it is to 1, the stronger the model's fitting ability is, the better the model explains the dependent variable, and the more stable the model is. RMSE is an indicator used to measure the deviation between the estimated value and the actual value. It can quantify the degree of difference between the predicted value and the actual value. The smaller the value, the closer the estimated value is to the actual value, and the higher the accuracy of the model.

[0201]

[0202] Among them, y' j represents the predicted value; represents the average value of the measured values; j represents the measured value; n represents the sample size;

[0203] 210. Based on the confusion matrix method, the water quality category is evaluated by using the water sample images of the verification points through the overall classification accuracy, recall rate and precision;

[0204] The spectral similarity evaluation method is used to match the water quality category, and the accuracy is evaluated according to the accuracy of the matching water sample category, including the following indicators: the overall classification accuracy is used to evaluate the proportion of the number of correctly classified samples in all samples, indicating the overall accuracy of the classification result, the recall rate is used to evaluate the proportion of the total number of real samples of this category, indicating the probability that the reference sample is correctly identified, and the precision rate is used to evaluate the ratio of the number of samples correctly classified into a certain category to the total number of samples classified into this category, reflecting the probability that the predicted sample is consistent with the actual sample category. Different matching algorithms are scored, where 2 points are assigned for successful spectral curve matching and 1 point is assigned for successful water quality category matching. The test samples are scored and sorted to screen the algorithm model suitable for water quality category evaluation.

[0205] The overall classification accuracy includes:

[0206]

[0207] The recall rate includes:

[0208]

[0209] The accuracy includes:

[0210]

[0211] Among them, TP represents a positive sample predicted to be positive; TN represents a negative sample predicted to be negative; FP represents a negative sample predicted to be positive; and FN represents a positive sample predicted to be negative.

[0212] The above describes the rapid water quality inversion method in the embodiment of the present invention. The following describes the rapid water quality inversion device in the embodiment of the present invention. Figure 5, an embodiment of the rapid water quality inversion device in the embodiment of the present invention includes:

[0213] The data acquisition module 301 is used to acquire water quality site monitoring data, drone remote sensing data and satellite remote sensing data based on a high frequency that meets preset requirements;

[0214] In this embodiment, the data resources of water quality monitoring stations are fully utilized to quickly respond to point problems in river and lake water quality. In order to ensure the timeliness and accuracy of the data, high-frequency data acquisition is performed, and water quality section monitoring station data is collected every 4 hours. High-frequency data updates can grasp the water quality status in real time, provide strong support for the timely discovery and treatment of water quality problems, and provide data support for subsequent multi-source data matching.

[0215] The data from water quality monitoring stations include key water quality indicators such as permanganate index, ammonia nitrogen, and total phosphorus. These indicators can intuitively reflect the degree of pollution and nutritional status of the water body. At the same time, parameters such as pH value, dissolved oxygen, water temperature, turbidity, and conductivity are obtained, which helps to fully understand the chemical and physical properties of the water body. In addition, the instantaneous flow velocity, instantaneous flow rate, and cross-sectional area data obtained help to more deeply understand the flow characteristics and hydrodynamic conditions of the water body. Through the data acquisition and monitoring method proposed in this patent, it is possible to more accurately grasp the changes in the water quality of rivers and lakes, and to promptly discover and deal with water quality problems from point to surface.

[0216] A near-synchronous remote sensing data source acquisition module 302 is used to perform time series analysis on the acquired water quality site monitoring data, and to match and acquire near-synchronous remote sensing data sources based on the acquired UAV remote sensing data and satellite remote sensing data;

[0217] In this embodiment, the continuity and real-time characteristics of the water quality station monitoring data are fully utilized, and the near-synchronous remote sensing data source is accurately matched and obtained by performing time series analysis on the water quality station monitoring data. In order to achieve near-synchronous remote sensing data source acquisition, the matching principle is to ensure that the acquisition time of the remote sensing image is as close as possible to the acquisition time of the water quality station data.

[0218] In this embodiment, the spatial resolution and spectral characteristics of the remote sensing image are considered simultaneously, and appropriate remote sensing satellites and sensors are selected to obtain remote sensing images with high resolution and high spectral information to ensure their matching and reliability with the water quality site data.

[0219] A preprocessing module 303 is used to preprocess the acquired near-synchronous remote sensing data source to obtain a preprocessed near-synchronous remote sensing data source;

[0220] In this embodiment, the remote sensing images are subjected to radiometric calibration and atmospheric correction batch preprocessing. Among them, radiometric calibration is one of the key steps in remote sensing image preprocessing, which aims to eliminate the performance differences of the sensor itself and the influence of the external environment on the image data, ensure the accuracy and reliability of the data, and enable it to truly reflect the reflective characteristics of the ground objects. Among them, atmospheric correction is another indispensable preprocessing step. Due to the presence of various factors in the atmosphere (such as scattering, absorption, etc.), remote sensing images will be subject to varying degrees of interference during transmission. In order to eliminate these interferences, an atmospheric correction algorithm based on the 6S model is used to batch process the images to restore the true reflectivity of the ground objects.

[0221] The spectral data acquisition module 304 is used to acquire the sites within the coverage of the near-synchronous remote sensing data source based on the pre-processed near-synchronous remote sensing data source, locate and extract the spectral data in the corresponding near-synchronous remote sensing data source based on the location information of the current site;

[0222] In this embodiment, the water quality stations located within the image coverage are determined based on the four-axis coordinate range information of the image, and the longitude and latitude coordinates of the station data within the range are converted into image row and column coordinates according to the RPC model to extract the spectral band information.

[0223] The spectrum library data set construction module 305 is used to construct the spectrum library data set based on the spectrum data in the near-synchronous remote sensing data source, the longitude and latitude of the station, the section name and the water quality parameters;

[0224] A water quality inversion model training module 306 is used to construct a water quality inversion model and train the water quality inversion model using the constructed spectral library data set to obtain a trained water quality inversion model;

[0225] In this embodiment, the water quality inversion model is constructed based on key characteristic variables that meet preset requirements;

[0226] The water quality inversion model is based on machine learning including random forest regression, support vector machine regression or XGBoost regression to achieve water quality parameter estimation of water bodies;

[0227] The key characteristic variables that meet the preset requirements include:

[0228] Extract the single-band reflectance data of remote sensing images corresponding to each sampling point, select any two bands for band combination, and obtain the potential characteristic variables related to the concentration of water quality parameters;

[0229] The potential characteristic variables are analyzed for correlation to obtain the key characteristic variables;

[0230] The potential characteristic variables are subjected to correlation analysis to obtain key characteristic variables, including: performing Pearson correlation analysis on the measured data of each water quality parameter and band combination to determine the key characteristic variables for constructing a water quality parameter inversion model.

[0231] The water quality inversion module 307 is used to perform water quality inversion using the trained water quality inversion model.

[0232] A water quality classification module 308 is used to classify water quality based on a spectral angle matching method;

[0233] In this embodiment, a spectral library of known water quality categories is constructed using different remote sensing data sources and water quality site monitoring data;

[0234] The current water quality category spectral library is used as the reference spectral library;

[0235] The spectral vector of each pixel in the remote sensing image is compared with the reference spectral vector, and the similarity between the spectra is measured by calculating the angle between the two spectral vectors, and the water quality category is classified based on the similarity;

[0236] The angle between the two spectral vectors is:

[0237]

[0238] Among them, θ x,y represents the angle between the reference spectrum vector X and the spectrum vector Y to be measured; n represents the number of bands of the spectrum curve; Xi represents the spectral reflectance value of the reference spectrum vector in the i-th band; Yi represents the spectral reflectance value of the spectrum vector to be measured in the i-th band.

[0239] Model evaluation module 309 is used to use the determination coefficient R 2 The water quality inversion model was evaluated by the root mean square error (RMSE);

[0240] Using the determination coefficient R 2 Measure the fit of the water quality inversion model;

[0241] The root mean square error RMSE is used to measure the deviation between the estimated value and the actual value;

[0242]

[0243] Among them, y' j represents the predicted value; represents the average value of the measured values; j represents the measured value; n represents the sample size;

[0244] The water quality category evaluation module 310 is used to evaluate the water quality category by using the water sample images of the verification points based on the confusion matrix method through the overall classification accuracy, recall rate and precision rate;

[0245] The overall classification accuracy includes:

[0246]

[0247] The recall rate includes:

[0248]

[0249] The accuracy includes:

[0250]

[0251] Among them, TP represents a positive sample predicted to be positive; TN represents a negative sample predicted to be negative; FP represents a negative sample predicted to be positive; and FN represents a positive sample predicted to be negative.

[0252] above Figure 3 The rapid water quality inversion device in the embodiment of the present invention is described in detail from the perspective of modular functional entities, and the electronic device in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0253] Figure 6 7 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. The electronic device 700 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 710 (for example, one or more processors) and a memory 720, and one or more storage media 730 (for example, one or more mass storage devices) storing application programs 733 or data 732. Among them, the memory 720 and the storage medium 730 can be temporary storage or permanent storage. The program stored in the storage medium 730 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the electronic device 700. Furthermore, the processor 710 can be configured to communicate with the storage medium 730 to execute a series of instruction operations in the storage medium 730 on the electronic device 700.

[0254] The electronic device 700 may also include one or more power supplies 740, one or more wired or wireless network interfaces 750, one or more input and output interfaces 750, and / or one or more operating systems 731, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. It will be appreciated by those skilled in the art that Figure 6The structure of the electronic device shown does not constitute a limitation on the electronic device, and may include more or less components than shown in the figure, or combine some components, or arrange the components differently.

[0255] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the rapid water quality inversion method.

[0256] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, device, or unit can refer to the corresponding process in the aforementioned method embodiment and will not be repeated here.

[0257] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0258] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A rapid water quality inversion method, characterized in that: include: Access high-frequency water quality site monitoring data, drone remote sensing data, and high-resolution satellite remote sensing data; Conduct time series analysis on the acquired water quality site monitoring data, and match and acquire near-synchronous remote sensing data sources based on the image data time of the acquired UAV remote sensing data and satellite remote sensing data; Preprocessing the acquired near-synchronous remote sensing data source to obtain a preprocessed near-synchronous remote sensing data source; Based on the preprocessed near-synchronous remote sensing data source, the stations within the coverage of the near-synchronous remote sensing data source are obtained, and the spectral data in the corresponding near-synchronous remote sensing data source are located and extracted based on the location information of the current station; A spectral library dataset is constructed based on the spectral data, longitude and latitude of the station, section name and water quality parameters in the near-synchronous remote sensing data source; A water quality inversion model is constructed, and the water quality inversion model is trained using the constructed spectral library data set to obtain a trained water quality inversion model; Using the trained water quality inversion model to perform water quality inversion; Among them, the water quality online monitoring site data includes: permanganate index, ammonia nitrogen, total phosphorus water quality indicators, pH value, dissolved oxygen, water temperature, turbidity, conductivity parameters, instantaneous flow velocity, instantaneous flow and cross-sectional area data.

2. The rapid water quality inversion method according to claim 1, characterized in that: The time series analysis of the acquired water quality site monitoring data is performed, and based on the image data time of the acquired UAV remote sensing data and satellite remote sensing data, the near-synchronous remote sensing data source is matched and acquired, including: Perform time series analysis on the acquired water quality site monitoring data; obtain the water quality site monitoring data collection time closest to the image data time, and obtain the corresponding water quality site monitoring data based on the closest water quality site monitoring data collection time; build a near-synchronous remote sensing data source based on the UAV remote sensing data and satellite remote sensing data corresponding to the current image data time.

3. The rapid water quality inversion method according to claim 1, characterized in that: The preprocessing of the acquired near-synchronous remote sensing data source to obtain the preprocessed near-synchronous remote sensing data source includes: The acquired near-synchronous remote sensing data source is subjected to radiation calibration and atmospheric correction processing to obtain a pre-processed near-synchronous remote sensing data source; The radiation calibration is to convert the original DN value of remote sensing impact into apparent reflectance or absolute radiation brightness, so as to achieve standardization and accuracy processing of image data; The atmospheric correction is that the 6S model simulates the surface reflectivity based on geometric parameters, atmospheric mode, aerosol mode, optical thickness, target altitude and sensor altitude, and compares the simulated surface reflectivity with the actual remote sensing data to perform atmospheric correction.

4. The rapid water quality inversion method according to claim 1, characterized in that: The method of acquiring a site within the coverage of the near-synchronous remote sensing data source based on the preprocessed near-synchronous remote sensing data source, and locating and extracting spectral data in the corresponding near-synchronous remote sensing data source based on the location information of the current site includes: According to the four-axis coordinate range information in the near-synchronous remote sensing data source, the stations within the coverage range of the near-synchronous remote sensing data source are judged; according to the rational polynomial coefficients provided in the remote sensing image data, the forward projection expression in the RPC model is used to convert the longitude and latitude coordinates of the stations within the coverage range into row and column coordinates on the remote sensing image data, and the spectral data is extracted based on the row and column coordinates on the remote sensing image.

5. The rapid water quality inversion method according to claim 1, characterized in that: The constructing of the water quality inversion model comprises: constructing the water quality inversion model based on key characteristic variables that meet preset requirements; The water quality inversion model is based on machine learning including random forest regression, support vector machine regression or XGBoost regression to achieve water quality parameter estimation of water bodies; The key characteristic variables that meet the preset requirements include: Extract the single-band reflectance data of remote sensing images corresponding to each sampling point, select any two bands for band combination, and obtain the potential characteristic variables related to the concentration of water quality parameters; Conduct correlation analysis on potential characteristic variables to obtain key characteristic variables; The potential characteristic variables are subjected to correlation analysis to obtain key characteristic variables, including: performing Pearson correlation analysis on the measured data of each water quality parameter and band combination to determine the key characteristic variables for constructing a water quality parameter inversion model.

6. The rapid water quality inversion method according to claim 1, characterized in that: The method further comprises: implementing water quality classification based on spectral angle matching method; A spectral library of known water quality categories was constructed using different remote sensing data sources and water quality site monitoring data; The current water quality category spectral library is used as the reference spectral library; Compare the spectral vector of each pixel in the remote sensing image with the reference spectral vector, and measure the similarity between the spectra by calculating the angle between the two spectral vectors; Classification of water quality categories is achieved based on the calculated similarity; The angle between the two spectral vectors is: Among them, θ x,y represents the angle between the reference spectrum vector X and the spectrum vector Y to be measured; n represents the number of bands of the spectrum curve; Xi represents the spectral reflectance value of the reference spectrum vector in the i-th band; Yi represents the spectral reflectance value of the spectrum vector to be measured in the i-th band.

7. The rapid water quality inversion method according to claim 6, characterized in that: The method further comprises: using the coefficient of determination R 2 The water quality inversion model was evaluated by the root mean square error (RMSE); Using the determination coefficient R 2 Measure the fit of the water quality inversion model; The root mean square error RMSE is used to measure the deviation between the estimated value and the actual value; Among them, y' j represents the predicted value; represents the average value of the measured values; j represents the measured value; n represents the sample size; Based on the confusion matrix method, the water quality category was evaluated by using the water sample images of the verification points through the overall classification accuracy, recall rate and precision rate. The overall classification accuracy includes: The recall rate includes: The accuracy includes: Among them, TP represents a positive sample predicted to be positive; TN represents a negative sample predicted to be negative; FP represents a negative sample predicted to be positive; and FN represents a positive sample predicted to be negative.

8. A rapid water quality inversion device, characterized in that: include: Data acquisition module, used to acquire high-frequency water quality site monitoring data, drone remote sensing data, and high-resolution satellite remote sensing data; The near-synchronous remote sensing data source acquisition module is used to perform time series analysis on the acquired water quality site monitoring data, and match and acquire the near-synchronous remote sensing data source based on the image data time of the acquired UAV remote sensing data and satellite remote sensing data; A preprocessing module is used to preprocess the acquired near-synchronous remote sensing data source to obtain a preprocessed near-synchronous remote sensing data source; The spectral data acquisition module is used to acquire the sites within the coverage of the near-synchronous remote sensing data source based on the pre-processed near-synchronous remote sensing data source, and locate and extract the spectral data in the corresponding near-synchronous remote sensing data source based on the location information of the current site; A spectral library dataset construction module is used to construct a spectral library dataset based on the spectral data in the near-synchronous remote sensing data source, the longitude and latitude of the station, the section name, and the water quality parameters; A water quality inversion model training module is used to construct a water quality inversion model and train the water quality inversion model using the constructed spectral library data set to obtain a trained water quality inversion model; A water quality inversion module is used to perform water quality inversion using the trained water quality inversion model; Among them, the water quality online monitoring site data includes: permanganate index, ammonia nitrogen, total phosphorus water quality indicators, pH value, dissolved oxygen, water temperature, turbidity, conductivity parameters, instantaneous flow velocity, instantaneous flow and cross-sectional area data.

9. An electronic device, comprising a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instructions in the memory to enable the electronic device to execute each step of the rapid water quality inversion method as described in any one of claims 1-7.

10. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the various steps of the rapid water quality inversion method as described in any one of claims 1-7 are implemented.

Citation Information

Cited By

  • Lake and reservoir total phosphorus and total nitrogen concentration inversion method and system based on Sentinel-2 image data

    CN119323705A

  • Method and system for improving satellite remote sensing inversion precision by fusing hyperspectral proximity

    CN119510359A

  • Water quality parameter inversion method based on hyperspectral data analysis

    CN119600454A

  • Full-coverage daily scale AOD inversion method, device and system and storage medium

    CN119939266A

  • Machine learning ammonia nitrogen inversion method and device, terminal and storage medium

    CN120009488A