Water quality remote sensing monitoring method, system and device and storage medium

Abstract

The embodiment of the invention provides a water quality remote sensing monitoring method, system and device and a storage medium. The method comprises the following steps: acquiring to-be-detected data; inputting the to-be-monitored data into a plurality of water body monitoring sub-models for monitoring to obtain a sub-model monitoring result of each water body monitoring sub-model corresponding to the to-be-monitored data; obtaining a training sample set; determining dynamic weights corresponding to the plurality of water body monitoring sub-models according to the training sample set, the to-be-monitored data, the monitoring results of the plurality of sub-models and a preset algorithm; and determining a water chlorophyll a concentration monitoring result of the target water area corresponding to the to-be-measured data based on the plurality of sub-model monitoring results and the dynamic weights corresponding to the plurality of water monitoring sub-models. The embodiment of the invention can effectively improve the accuracy of monitoring the water chlorophyll a concentration of the target water area, and can be widely applied to the field of water monitoring.

Description

Technical Field

[0001] The present invention relates to the field of water body monitoring, and in particular, to a water quality remote sensing monitoring method, system, device and storage medium. Background Art

[0002] Water bodies are an important part of the global ecosystem and play a key role in regulating the climate, providing ecological services, supporting biodiversity, and ensuring the safety of human water use. However, with the continuous intensification of industrialization and agricultural activities, water bodies worldwide are facing increasingly serious eutrophication problems. Eutrophication is mainly manifested by the increase in the concentration of nutrients such as nitrogen and phosphorus in water bodies, leading to abnormal algal blooms, ultimately affecting water quality, disrupting the aquatic ecological balance, and even triggering the phenomenon of water blooms. As the main photosynthetic pigment of phytoplankton, the concentration of chlorophyll a is widely used in water quality monitoring as an important biological indicator to measure the degree of water body eutrophication. Excessive chlorophyll a concentration not only causes hypoxia in water bodies and degradation of the ecological environment, but also may threaten the safety of drinking water. Therefore, accurately monitoring the chlorophyll a concentration in water bodies has important practical significance for water resource management and ecological environment protection.

[0003] However, the accuracy of the existing technology for monitoring the chlorophyll a concentration in water bodies of the target water area still needs to be improved. Summary of the Invention

[0004] In view of this, to solve one of the above problems, the purpose of the embodiments of the present invention is to provide a water quality remote sensing monitoring method, system, device and storage medium, which can effectively improve the accuracy of monitoring the chlorophyll a concentration in water bodies of the target water area.

[0005] In a first aspect, an embodiment of the present invention provides a water quality remote sensing monitoring method, including the following steps:

[0006] Obtain data to be measured; the data to be measured is determined according to water quality remote sensing monitoring data;

[0007] Input the data to be measured into a plurality of water body monitoring sub-models for monitoring, and obtain the sub-model monitoring results corresponding to each water body monitoring sub-model for the data to be measured;

[0008] Obtain a training sample set; determine the dynamic weights corresponding to the plurality of water body monitoring sub-models according to the training sample set, the data to be measured, the sub-model monitoring results of the plurality of water body monitoring sub-models, and a preset algorithm;

[0009] Determine the monitoring result of the chlorophyll a concentration in the water body of the target water area corresponding to the data to be measured based on the sub-model monitoring results of the plurality of water body monitoring sub-models and the dynamic weights corresponding to the plurality of water body monitoring sub-models.

[0010] Specifically, the training sample set includes an input data sample set and a monitoring result sample set; the dynamic weight corresponding to the water body monitoring sub-model is determined as follows:

[0011] Based on the data to be measured, the sub-model monitoring results corresponding to each of the water body monitoring sub-models for the data to be measured, the input data sample set, and the monitoring result sample set, calculate the weighted comprehensive distance corresponding to each of the water body monitoring sub-models;

[0012] Calculate the weighted comprehensive distances corresponding to each of the water body monitoring sub-models to obtain the model weighted comprehensive distance;

[0013] Based on the weighted comprehensive distance corresponding to the water body monitoring sub-model and the model weighted comprehensive distance, calculate to determine the dynamic weight corresponding to the water body monitoring sub-model.

[0014] Specifically, the weighted comprehensive distance corresponding to the water body monitoring sub-model is determined as follows:

[0015] Based on the sub-model monitoring results corresponding to the water body monitoring sub-model for the data to be measured and the monitoring result sample set, calculate to obtain a prediction error set;

[0016] Based on the prediction error set and a preset condition, make a judgment, and determine the neighborhood sample space corresponding to the data to be measured according to the result of the judgment, the input data sample set, and the monitoring result sample set;

[0017] Based on the data to be measured, the sub-model monitoring results corresponding to the water body monitoring sub-model for the data to be measured, and the neighborhood sample space corresponding to the data to be measured, calculate to determine the weighted comprehensive distance corresponding to the water body monitoring sub-model.

[0018] Specifically, the step of making a judgment based on the prediction error set and a preset condition, and determining the neighborhood sample space corresponding to the data to be measured according to the result of the judgment, the input data sample set, and the monitoring result sample set includes:

[0019] Perform weighted sorting on the error values in the prediction error set to obtain a weighted sorting result;

[0020] Based on the weighted sorting result and a preset condition, make a judgment, and construct the neighborhood sample space corresponding to the data to be measured based on several input data samples that meet the conditions and their corresponding monitoring result samples according to the result of the judgment.

[0021] Specifically, the step of determining the weighted comprehensive distance corresponding to the water body monitoring sub-model includes:

[0022] Calculate based on a first preset formula, the data to be measured, and the neighborhood sample space corresponding to the data to be measured, to obtain the feature-weighted Euclidean distance corresponding to the water body monitoring sub-model;

[0023] Calculate based on a second preset formula, the water body monitoring prediction result corresponding to the water body monitoring sub-model for the data to be measured, and the neighborhood sample space corresponding to the data to be measured, to obtain the label difference metric corresponding to the water body monitoring sub-model;

[0024] Calculate based on a third preset formula, a preset coefficient, and the feature-weighted Euclidean distance and label difference metric corresponding to the water body monitoring sub-model, to determine the weighted comprehensive distance corresponding to the water body monitoring sub-model.

[0025] Further, the method further includes:

[0026] Generate a spatial distribution map of the chlorophyll a concentration of the target water area according to the monitoring results of the chlorophyll a concentration of the water body in the target water area corresponding to each data to be measured;

[0027] Analyze the spatial distribution map of the chlorophyll a concentration of the target water area, and determine the water body monitoring result of the target water area according to the analysis result.

[0028] Further, the method further includes:

[0029] Obtain the measured results of the chlorophyll a concentration of the water body in the target water area corresponding to each data to be measured;

[0030] Calculate according to a fourth preset formula, based on the average value of the target water area monitoring results and the monitoring results and measured results of the chlorophyll a concentration of the water body in the target water area corresponding to each data to be measured, to obtain a first evaluation value; the average value of the target water area monitoring results is calculated through the monitoring results of the chlorophyll a concentration of the water body in the target water area corresponding to each data to be measured;

[0031] Calculate according to a fifth preset formula, based on the monitoring results and measured results of the chlorophyll a concentration of the water body in the target water area corresponding to each data to be measured, to obtain a second evaluation value;

[0032] Analyze based on the first evaluation value and the second evaluation value to determine the water body monitoring accuracy of the target water area.

[0033] On the other hand, an embodiment of the present invention further provides a water quality remote sensing monitoring system, including:

[0034] A first module, configured to obtain data to be measured; the data to be measured is determined according to water body remote sensing monitoring data;

[0035] A second module, configured to input the data to be measured into a plurality of water body monitoring sub-models for monitoring, and obtain the sub-model monitoring results corresponding to the data to be measured for each of the water body monitoring sub-models;

[0036] A third module, configured to obtain a training sample set and determine dynamic weights corresponding to the plurality of water body monitoring sub-models according to the training sample set, the data to be measured, the monitoring results of the plurality of sub-models, and a preset algorithm;

[0037] A fourth module, configured to determine the monitoring result of the chlorophyll-a concentration of the water body in the target water area corresponding to the data to be measured based on the monitoring results of the plurality of sub-models and the dynamic weights corresponding to the plurality of water body monitoring sub-models.

[0038] On the other hand, an embodiment of the present invention further provides a water quality remote sensing monitoring device, including:

[0039] At least one processor;

[0040] At least one memory, configured to store at least one program;

[0041] When the at least one program is executed by the at least one processor, the at least one processor implements the method as described above.

[0042] On the other hand, an embodiment of the present invention further provides a computer-readable storage medium, in which a program executable by a processor is stored, and the program executable by the processor is used to execute the method as described above when executed by the processor.

[0043] In summary, implementing the embodiments of the present invention includes the following beneficial effects:

[0044] The embodiment of the present invention provides a water quality remote sensing monitoring method, system, device and storage medium. The method calculates by inputting the data to be measured into a plurality of water body monitoring sub-models, and determines the dynamic weights corresponding to each water body monitoring sub-model according to the obtained sub-model monitoring results and the training sample set. Finally, the weighted calculation is performed on the sub-model monitoring results to obtain the monitoring result of the chlorophyll-a concentration of the water body in the target water area corresponding to the data to be measured. In this regard, the method provided by the present invention determines the dynamic weights of the sub-models and performs weighted summation calculation on the sub-model monitoring results, and determines the final monitoring result according to the calculation result. Such a monitoring process can integrate the algorithm advantages of each water body monitoring sub-model, enhance the generalization ability and stability of the water quality remote sensing monitoring method, and improve the accuracy of monitoring the chlorophyll-a concentration of the water body in the target water area. Description of the Drawings

[0045] Figure 1 It is a schematic flowchart of the steps of a water quality remote sensing monitoring method provided by an embodiment of the present invention;

[0046] Figure 2 It is a schematic diagram of the step flow of an embodiment of a water quality remote sensing monitoring method provided by an embodiment of the present invention;

[0047] Figure 3 It is a structural block diagram of a water quality remote sensing monitoring system provided by an embodiment of the present invention;

[0048] Figure 4 It is a structural block diagram of a water quality remote sensing monitoring device provided by an embodiment of the present invention. Specific embodiments

[0049] The following further elaborates on the present invention in detail in conjunction with the accompanying drawings and specific embodiments. For the step numbers in the following embodiments, they are only set for the convenience of elaboration and explanation, and no limitation is imposed on the order between the steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0050] Explanations for several terms involved in this application are as follows:

[0051] Ensemble Machine-Deep Learning (EMD): An ensemble method that combines traditional machine learning and deep learning models, aiming to improve prediction accuracy and robustness by integrating the advantages of multiple models. It combines traditional machine learning algorithms (such as random forest, support vector machine, etc.) with deep learning networks (such as CNN, RNN, etc.), thus achieving better performance when dealing with complex and non-linear data. EMD can effectively handle various tasks, reduce overfitting, and enhance the generalization ability of the model. Especially in large-scale datasets and multi-task learning, it can significantly improve accuracy and stability.

[0052] Support Vector Machine (SVM): A machine learning algorithm. It divides data points of different categories by finding the optimal hyperplane. In water body monitoring, SVM can be used for classification tasks, such as the detection of water quality pollutants and the classification of water body types. By converting water quality monitoring data (such as temperature, pH value, dissolved oxygen, etc.) into a high-dimensional feature space, SVM can effectively identify the presence of pollutants or changes in the water body, with high classification accuracy and strong generalization ability. The advantage of SVM is that it can handle non-linear problems and adapt to different water quality changes, which helps to monitor the changes in the water environment in real time.

[0053] Random Forest (RF): An ensemble learning method that improves the prediction accuracy and robustness of a model by constructing multiple decision trees and performing voting or averaging. Each decision tree randomly samples the data during training and introduces randomness in feature selection, thus reducing the overfitting risk of a single tree. Random Forest has strong generalization ability, can handle large-scale datasets, and performs excellently in classification and regression problems. Due to its simplicity, ease of understanding, and high computational efficiency, Random Forest is widely used in data mining, image recognition, financial risk prediction, etc. Especially when dealing with feature selection and high-dimensional data, it demonstrates good performance.

[0054] Adaptive Boosting (AdaBoost): An ensemble learning method that improves the prediction accuracy of a model by combining multiple weak classifiers into a strong classifier. Its basic idea is to gradually adjust the weights of each weak classifier so that subsequent classifiers pay more attention to the data points misclassified by the previous classifier. AdaBoost combines the results of each weak classifier through weighted voting, thus significantly improving the classification performance. This algorithm has strong anti-noise ability and low computational complexity, and is widely used in pattern recognition, image classification, text processing, etc. Due to its simplicity and high efficiency, AdaBoost is often used in practical applications to improve the accuracy and robustness of classifiers, especially when dealing with imbalanced data.

[0055] Multi-Layer Perceptron (MLP): A feedforward neural network composed of an input layer, one or more hidden layers, and an output layer. Each layer of neurons is connected to all neurons in the previous layer, and uses non-linear activation functions (such as ReLU, Sigmoid, etc.) to process information. MLP adjusts the weights through the backpropagation algorithm, enabling the network to continuously optimize during training to achieve the mapping between input and output. It can handle complex pattern recognition tasks and is widely used in image recognition, speech processing, natural language processing, etc.

[0056] Two-Band Index (2BA): An index used for remote sensing image analysis. It usually enhances the ability to distinguish ground object features by combining reflectance data from two different bands. Common two-band indices include NDVI (Normalized Difference Vegetation Index), etc. Its basic principle is that through the reflectance differences in different bands, it can effectively distinguish ground object categories such as vegetation, water bodies, and soil. The two-band index is simple and efficient, widely used in remote sensing image processing, especially in fields such as agriculture, forestry, and environmental monitoring, and can quickly evaluate important information such as plant health status and soil moisture. By adjusting the combination method of different bands, the 2BA method can be optimized according to specific needs to improve the accuracy of data processing.

[0057] Blue-Green Index (BG): A remote sensing image index mainly used to distinguish the features of water bodies from those of vegetation, soil, etc. By utilizing the reflectance differences between the blue band and the green band, this index can effectively identify the distribution and water quality status of water bodies. When the value of the blue-green index is usually high, it indicates a water body area, and when it is low, it indicates a non-water body area. The BG index has important applications in water resource management, environmental monitoring, and ecological protection, especially in water area remote sensing analysis, flood monitoring, and water pollution assessment, and can provide accurate water area information.

[0058] Red-Green Index (RG): A remote sensing image index that mainly analyzes ground object features by comparing the reflectance differences between the red band and the green band. This index is usually used for vegetation monitoring and can effectively distinguish vegetation areas from non-vegetation areas. When the value of the red-green index is high, it usually indicates a vegetation area, while a low value indicates other ground objects such as bare soil or water bodies. The RG index is simple to calculate and widely used in fields such as agriculture, forestry, and ecological environment monitoring to help evaluate vegetation health status and land use changes.

[0059] Fluorescence Line Height Blue Index (FLH Blue): An index used for remote sensing image analysis that mainly evaluates the characteristics of water bodies or vegetation by analyzing the fluorescence reflectance in the blue band. This index focuses on the fluorescence signal intensity in the blue band and can effectively identify information such as the algae content in water bodies and the plant growth status. FLH Blue has important applications in water quality monitoring, ecological environmental protection, and marine research, especially in evaluating water body eutrophication and ecological health status, providing important remote sensing data support.

[0060] Be16NDPhyI: A remote sensing index used in water body monitoring, typically for evaluating the physical properties of water bodies, especially in large-scale water area analysis through satellite imagery. This index combines reflectance information from different bands, helping to detect water quality, pollutants, suspended solids, and changes in water areas. Through Be16NDPhyI, researchers can more accurately identify the degree of water pollution and water quality status, providing data support for water resource management and environmental protection. This method is applicable to large-scale water body monitoring, such as lakes, rivers, and oceans.

[0061] Normalized Difference Turbidity Index (NDTI): A remote sensing index mainly used for monitoring water turbidity. NDTI evaluates the concentration of suspended solids and turbidity in water by utilizing the reflectance difference between the red and near-infrared bands in remote sensing imagery. This index is highly sensitive to water pollution, sediment content, etc., and is widely used in water quality monitoring, environmental assessment of lakes and rivers, helping scientific researchers and environmental management departments to promptly detect water quality problems and take corresponding measures for treatment.

[0062] Blue-Red Ratio Index (BR): An index based on remote sensing data for analyzing and evaluating the characteristics of vegetation and water bodies. It reflects the surface cover type and ecological environment status by calculating the reflectance ratio of the blue and red bands in remote sensing imagery. This index can effectively distinguish different types of vegetation, soil, and water bodies, and is widely used in fields such as agriculture, forestry, and water quality monitoring. Through the BR value, researchers can monitor and analyze vegetation growth, health status, and water quality changes.

[0063] Green-Red Ratio Index (GR): A remote sensing index mainly used for evaluating the growth status and health of vegetation. It reflects the photosynthesis and biomass distribution of surface vegetation by calculating the reflectance ratio of the green band to the red band in remote sensing imagery. The GR index can effectively distinguish healthy vegetation from other ground objects, such as bare soil and water bodies, and is commonly used in agricultural monitoring, ecological assessment, and environmental monitoring. Through the GR value, the growth trend of vegetation can be tracked in real time, providing a scientific basis for agricultural management and ecological protection.

[0064] Ratio Vegetation Index (RVI): A remote sensing index used to evaluate the growth status of vegetation around water bodies. RVI reflects the health of vegetation by calculating the ratio of the reflectance in the near-infrared band to that in the red band. In water environment monitoring, RVI can help identify the vegetation cover around water bodies and analyze the impact of water bodies on the surrounding ecological environment. This index is widely used in ecological monitoring around water areas such as rivers, lakes, and wetlands, providing an effective means of monitoring the growth changes of vegetation around water bodies.

[0065] Yusense Map software: A Geographic Information System (GIS) software mainly used for data visualization, spatial analysis, and map making. It supports various geospatial data formats, can process and analyze geographic information data such as remote sensing images, terrain data, and climate data. Yusense Map has a user-friendly interface, provides rich map drawing and data analysis tools, and is widely used in fields such as environmental monitoring, urban planning, and agricultural management. This software helps users efficiently manage geographic data, make analysis and decisions, and display spatial information, improving work efficiency and accuracy.

[0066] Ratio Spectral Index (RSI): An index based on remote sensing data mainly used to analyze the spectral characteristics of surface materials. It highlights the spectral characteristics of specific ground objects or phenomena by calculating the ratio of reflectance between different bands. RSI is widely used in fields such as vegetation monitoring, soil analysis, and water body monitoring, and can effectively distinguish different types of ground objects, identify the health status of vegetation, land use changes, etc. Through the calculation of RSI, it can help researchers conduct more accurate analysis and evaluation of the ecological environment, agriculture, and natural resources.

[0067] Greenness Chlorophyll Index (GCI): A remote sensing index used to evaluate the chlorophyll content of phytoplankton in water bodies. Through the difference in reflectance between the green band and the near-infrared band in remote sensing data, GCI can effectively reflect the concentration and growth status of algae in water quality. This index is widely used in water quality monitoring of lakes, reservoirs, and coastal waters, helps judge the degree of water eutrophication, monitor algal blooms, and provides a scientific basis for water resource management and ecological protection. GCI has high sensitivity and helps to monitor water environment changes in real time.

[0068] Normalized Difference Vegetation Index (NDVI): A remote sensing index mainly used to evaluate the growth status of vegetation around water areas. NDVI is calculated based on the reflectance difference between the near-infrared band and the red band, reflecting the health of vegetation. The growth status of vegetation around water bodies is closely related to water quality, and NDVI can help monitor phenomena such as water eutrophication and algal blooms. This index has wide applications in water environment monitoring and can provide a scientific basis for water quality improvement, ecological protection, and water resource management.

[0069] Normalized Difference Water Index (NDWI) is an index used for water body monitoring in remote sensing images. It extracts water body information by comparing the reflectance of the near-infrared band and the short-wave infrared band. The NDWI value is usually between -1 and +1. The closer the value is to +1, the more obvious the water body is, while a negative value represents a non-water body area. NDWI is widely used in fields such as water resource monitoring, flood monitoring, and analysis of changes in lakes and rivers.

[0070] Correlation coefficient (R 2 , coefficient of determination): A statistical index that measures the strength of the linear relationship between two variables, with a value range from 0 to 1. The closer the R 2 value is to 1, the stronger the relationship between the two variables and the better the fitting effect of the model. It is usually used in regression analysis to help evaluate the explanatory power of the independent variable for the dependent variable. In fields such as water body monitoring, climate research, and economic analysis, R 2 is used to judge the accuracy and reliability of the prediction model. The higher the R 2 value means that the model can more effectively predict and explain the changes in the data.

[0071] Root Mean Square Error (RMSE) is a commonly used index that measures the difference between the predicted values and the actual values of a model. It calculates the square of the difference between the predicted values and the true values, takes the average, and then takes the square root, reflecting the magnitude of the model prediction error. The smaller the RMSE value, the higher the accuracy of the model prediction and the smaller the error. In fields such as water body monitoring and weather forecasting, RMSE is widely used to evaluate the performance of the model, helping to improve the prediction model and optimize the decision-making process.

[0072] As Figure 1 shown, the embodiments of the present invention provide a water quality remote sensing monitoring method, and the steps S100 to S400 included therein are as follows.

[0073] S100: Obtain the data to be measured; the data to be measured is determined according to the water body remote sensing monitoring data.

[0074] In this embodiment, first, a multi - spectral sensor carried on a remote sensing platform is used to perform multi - spectral imaging on the target water body area to obtain surface reflectance data, providing high - resolution input information for subsequent analysis. Second, based on the obtained surface reflectance data, multiple spectral indices are calculated to enhance the features related to the chlorophyll - a concentration in the water body. Finally, the corresponding surface reflectance and spectral indices on the remote sensing image are used as the data to be measured for subsequent water body monitoring calculations.

[0075] Specifically, in the process of obtaining the data to be measured, the embodiment of the present invention will perform pre - processing on the multi - spectral image to obtain a multi - spectral pre - processed image; the pre - processing methods include inter - channel registration, mosaicking, and radiometric correction. Second, based on the surface reflectance data and the multi - spectral pre - processed image, the multi - spectral band range corresponding to the multi - spectral pre - processed image is determined. Calculations are performed based on the corresponding multi - spectral band range and several preset formulas to obtain the spectral indices corresponding to the multi - spectral image.

[0076] In some embodiments, the multi - spectral sensor carried on the remote sensing platform includes a blue - light band (B), a green - light band (G), a red - light band (R), a red - edge band (RE), and a near - infrared band (NIR). Specifically: the wavelength of the blue - light band is 450 nm, and the bandwidth is 35 nm; the wavelength of the green - light band is 555 nm, and the bandwidth is 27 nm; the wavelength of the red - light band is 660 nm, and the bandwidth is 22 nm; the wavelength of the red - edge band is 720 nm, and the bandwidth is 10 nm; the wavelength of the near - infrared band is 840 nm, and the bandwidth is 30 nm.

[0077] Spectral indices play a key role in the analysis based on multi - spectral images, especially in the inversion of chlorophyll - a concentration in water bodies. These indices can sensitively reflect the optical properties and biochemical states of water bodies, providing core spectral information support for the inversion model. In response to the need for spectral sensitivity in the inversion of chlorophyll - a concentration in water bodies, the present invention synthesizes the research results of predecessors, combines the band characteristics of multi - spectral images and the optical characteristics of water bodies, and screens out spectral indices closely related to the chlorophyll - a concentration in water bodies, including but not limited to two - band index (2BA), blue - green index (BG), red - green index (RG), fluorescence line height blue index (FLH Blue), Be16NDPhyI, normalized difference turbidity index (NDTI), blue - red ratio index (BR), green - red ratio index (GR), ratio vegetation index (RVI), ratio spectral index (RSI), chlorophyll index (GCI), and normalized difference vegetation index (NDVI). These indices cover different band combination relationships, can effectively characterize the chlorophyll content in water bodies and its dynamic changes, laying a scientific foundation and data support for model development and accuracy improvement.

[0078] Specifically, the selected spectral indices and their calculation expressions are as follows:

[0079] Two-Band Index (2BA), the calculation expression is RE / R;

[0080] Blue-Green Index (BG), the calculation expression is G / B;

[0081] Red-Green Index (RG), the calculation expression is R / G;

[0082] Fluorescence Line Height Blue Index (FLH Blue), the calculation expression is G - [R + (B - R)];

[0083] A remote sensing index (Be16NDPhyI), the calculation expression is R - [RE + (G - RE)];

[0084] Normalized Difference Turbidity Index (NDTI), the calculation expression is (R - G) / (R + G);

[0085] Blue-Red Ratio Index (BR), the calculation expression is B / R;

[0086] Green-Red Ratio Index (GR), the calculation expression is G / R;

[0087] Ratio Vegetation Index (RVI), the calculation expression is NIR / R;

[0088] Ratio Spectral Index (RSI), the calculation expression is R / B;

[0089] Greenness Chlorophyll Index (GCI), the calculation expression is (G / NIR) - 1;

[0090] Normalized Difference Water Index (NDWI), and its calculation formula is (G - NIR) / (G + NIR).

[0091] It should be noted that for the water quality remote sensing monitoring method provided in the embodiments of the present invention, the data source or data acquisition method is not limited to the forms presented in this article.

[0092] S200: Input the data to be measured into several water body monitoring sub-models for monitoring, and obtain the sub-model monitoring results corresponding to the data to be measured for each water body monitoring sub-model.

[0093] The processed data to be measured is respectively input into several water body monitoring sub-models for calculation to obtain the corresponding monitoring results of the sub-models, which are used as the data basis for subsequent water body monitoring calculations.

[0094] Specifically, the data to be measured x j is obtained from the data set to be measured determined by the water body monitoring data and input the data to be measured x j into the water body monitoring sub-model set (k is the number of water body monitoring sub-models, M k is the k-th water body monitoring sub-model), and several water body monitoring sub-models are used for calculation to obtain the sub-model monitoring results corresponding to the data to be measured for each water body monitoring sub-model, which is expressed as:

[0095]

[0096] Among them, is the sub-model monitoring result corresponding to the data to be measured for the k-th water body monitoring sub-model, k is the number of water body monitoring sub-models, and M k is the water body monitoring sub-model.

[0097] S300: Obtain the training sample set; according to the training sample set, the data to be measured, several sub-model monitoring results and a preset algorithm, determine the dynamic weights corresponding to several water body monitoring sub-models.

[0098] Introduce the training set (where represents the d-dimensional feature space, is the corresponding water body monitoring result (the corresponding label value)), and calculate based on the training sample set for the already obtained data basis to determine the dynamic weights corresponding to the data to be measured for each water body monitoring sub-model, that is, determine the correlation of each algorithm in the water body monitoring process corresponding to the data to be measured. The higher the correlation of the water body monitoring sub-model, the higher the corresponding dynamic weight, and the higher the referenceability of its sub-model monitoring result to the comprehensive monitoring result corresponding to the data to be measured.

[0099] S400: Determine the monitoring result of the chlorophyll-a concentration of the water body in the target water area corresponding to the data to be measured based on the monitoring results of several sub-models and the dynamic weights corresponding to several water body monitoring sub-models.

[0100] By performing weighted calculation on the monitoring results of the sub-models corresponding to the data to be measured, obtain the comprehensive monitoring result corresponding to the data to be measured.

[0101] Specifically, the monitoring result of the chlorophyll-a concentration of the water body in the target water area corresponding to the data to be measured Is calculated by the following formula:

[0102]

[0103] Wherein, ω k ′ is the dynamic weight corresponding to the k-th water body monitoring sub-model, Is the monitoring result of the sub-model corresponding to the k-th water body monitoring sub-model for the data to be measured, and K is the number of models of the water body monitoring sub-models.

[0104] In some embodiments, in step S300, the dynamic weight corresponding to the water body monitoring sub-model can be determined in the following manner:

[0105] S310: Based on the data to be measured, the monitoring results of the sub-models corresponding to each water body monitoring sub-model for the data to be measured, the input data sample set, and the monitoring result sample set, perform calculations to determine the weighted comprehensive distance corresponding to each water body monitoring sub-model.

[0106] Specifically, the training sample set Includes the input data sample set And the monitoring result sample set The weighted comprehensive distance corresponding to the water body monitoring sub-model is determined in the following manner:

[0107] S311: Based on the monitoring results of the sub-models corresponding to the water body monitoring sub-model for the data to be measured and the monitoring result sample set, perform calculations to obtain the prediction error set;

[0108] Specifically, the calculation of the prediction error is implemented by the following formula:

[0109]

[0110] Wherein, ε i Is the prediction error, y i Is the monitoring result sample, Is the monitoring result of the sub-model corresponding to the k-th water body monitoring sub-model for the data to be measured.

[0111] S312: Make a judgment based on the prediction error set and preset conditions, and determine the neighborhood sample space corresponding to the data to be measured according to the judgment result, the input data sample set, and the monitoring result sample set.

[0112] In some embodiments, step S312 can be implemented through the following steps:

[0113] Perform weighted sorting on the error values in the prediction error set to obtain a weighted sorting result;

[0114] Make a judgment according to the weighted sorting result and preset conditions, and construct the neighborhood sample space corresponding to the data to be measured based on several input data samples that meet the conditions and their corresponding monitoring result samples.

[0115] Specifically, construct the test sample x j by optimizing the following formula to obtain the neighborhood sample space

[0116]

[0117] where n represents the preset number of neighborhood samples, and ε i is the prediction error, and the subset S is from the training set

[0118] The core idea of this formula is to select the most relevant neighborhood samples by minimizing the prediction error between the training samples and the test sample. Among them: (1) Goal: Select a subset of the training set such that the sum of the errors of the samples in it is the smallest, that is, select the samples closest to the test sample; (2) Optimization problem: Select n samples in the training set to minimize the sum of the prediction errors ε j between them and the test sample x i ; (3) Dynamic weighted sorting: Determine which training samples are more important for the prediction of the test sample through weighted sorting, so as to select the most relevant neighborhood samples.

[0119] S313: Calculate based on the data to be measured, the sub-model monitoring result corresponding to the water body monitoring sub-model for the data to be measured, and the neighborhood sample space corresponding to the data to be measured, and determine the weighted comprehensive distance corresponding to the water body monitoring sub-model.

[0120] Specifically, to enhance local sample consistency, define the weighted comprehensive distance function D(x j between the test sample x and the neighborhood sample space , x j , x i ), which is composed of the feature-weighted Euclidean distance and the label difference metric.

[0121] In some embodiments, the process of determining the weighted comprehensive distance corresponding to the water body monitoring sub-model in step S313 can be implemented through the following steps:

[0122] a. Calculate based on the first preset formula, the data to be measured, and the neighborhood sample space corresponding to the data to be measured to obtain the characteristic weighted Euclidean distance corresponding to the water body monitoring sub-model;

[0123] Specifically, the characteristic weighted Euclidean distance d feat (x j , x i ) is determined by the following formula:

[0124]

[0125] Wherein, d represents the dimension of feature l; ∣x j,l -x i,l ∣ represents the difference between the data to be measured x j and the sample data x in the domain space i on feature l; w l >0 represents the normalized weight of feature l, satisfying The norm p is a fixed value that controls the non-linearity degree of the distance.

[0126] The above formula is used to calculate the characteristic weighted Euclidean distance between x j and x i , where the contribution of each feature is given different weights. The calculation process is as follows:

[0127] (1) For samples x j and sample x i , we first compare their differences ∣x j,l -x i,l ∣ on each feature, where l represents the index of the feature;

[0128] (2) Calculate the weighted value of each feature difference: Multiply each feature difference by the corresponding weight w l , where the weight w l is the normalized weight of feature l, and w l >0 and the sum of weights (ensuring that the weights of all features add up to 1);

[0129] (3) Sum up the weighted differences of all features;

[0130] (4) Take the absolute value of the summation result and normalize it using the norm p (p = 2 in the case of Euclidean distance) to obtain the characteristic weighted Euclidean distance d feat (x j , x i), which can adjust the influence of each feature according to the importance of the feature.

[0131] Specifically, the norm p is a fixed value that controls the degree of non-linearity of the distance, and it defines the norm type of the weighted Euclidean distance. The common cases are:

[0132] When p = 2, the formula becomes the classical Euclidean distance.

[0133] When p = 1, the formula becomes the Manhattan distance (L1 norm).

[0134] When p > 2, the measurement of the distance will give greater weight to larger differences, affecting the degree of non-linearity of the calculation.

[0135] By selecting different p values, the model can emphasize feature differences to different extents, adapting to different data distributions and problem requirements.

[0136] Specifically, d in the above formula represents the dimension of the feature, that is, the dimension of the feature space of the sample. That is to say, the sample x j and the sample x i both have d features, and the summation symbol represents weighted calculation for all features. Specifically, the formula weights the differences of each feature l and then sums them up to finally obtain an overall distance value. d is the number of feature quantities or dimensions of the data, reflecting the size of the feature space of the sample.

[0137] Specifically, the feature l in the above formula refers to each attribute or dimension in the sample data, that is, each individual feature of the input data. Specifically:

[0138] Assume that each sample x j is a vector composed of d features, expressed as:

[0139] x j =(x j,1 , x j,2 ,…, x j,d )

[0140] where x j,l represents the value of the sample x j on the l-th feature. l is the index of the feature, and the value range is l = 1, 2, …, d.

[0141] Feature l is each of these data attributes. For example, in remote sensing image data, Feature l may represent the red, green, and blue band values of pixels, or other features such as vegetation indices and humidity extracted from remote sensing images or environmental data. In summary, Feature l is each attribute or dimension in the sample data. It is each input variable in the dataset, usually various numerical values obtained based on remote sensing images, ground measurements, or other observations in certain environmental monitoring.

[0142] b. Calculate based on the second preset formula, the water body monitoring prediction result corresponding to the data to be measured of the water body monitoring sub-model, and the neighborhood sample space corresponding to the data to be measured, to obtain the label difference metric corresponding to the water body monitoring sub-model.

[0143] Specifically, the label difference metric d label (x j , x i ) is determined by the following formula:

[0144]

[0145] Among them, y i is the sample data of the monitoring result in the neighborhood sample space , is the sub-model monitoring result corresponding to the k-th water body monitoring sub-model for the data to be measured.

[0146] c. Calculate based on the third preset formula, the preset coefficient, and the feature-weighted Euclidean distance and label difference metric corresponding to the water body monitoring sub-model to determine the weighted comprehensive distance corresponding to the water body monitoring sub-model.

[0147] Specifically, the weighted comprehensive distance D(x j , x i ) of the water body monitoring sub-model consists of the feature-weighted Euclidean distance d feat (x j , x i ) and the label difference metric d label (x j , x i ) and is defined by the following formula through linear weighting:

[0148]

[0149] Among them, λ > 0 is the weight coefficient for adjusting the relative importance of the two parts of the distance.

[0150] Specifically, the weight coefficient λ is not a fixed value but an adjustable hyperparameter. Its role is to balance the feature-weighted Euclidean distance d feat (x j , x i) and the label difference metric d label (x j , x i ). According to different data and tasks, by selecting different λ values, the contribution degrees of the two parts of distance metrics to the overall distance calculation can be adjusted.

[0151] Usually, λ needs to be adjusted through methods such as cross - validation to select the value most suitable for the current task. For example, an overly high λ value may make the label difference metric dominate in the calculation, while an overly low λ value may lead to an overly large influence of the feature - weighted Euclidean distance. Therefore, λ is a parameter that needs to be determined during the model training process.

[0152] Specifically, the relative importance of distance in the formula refers to the contribution ratio of the feature distance and the label difference metric to the overall distance calculation. The weight coefficient λ determines the relative importance of the feature - weighted Euclidean distance d feat (x j , x i ) and the label difference metric d label (x j , x i ) in the final distance calculation.

[0153] When λ is large, the label difference metric d label (x j , x i ) has a greater impact on the final distance, that is, the model attaches more importance to the label differences between samples (such as the differences in chlorophyll a concentration); when λ is small, the feature - weighted Euclidean distance d feat (x j , x i ) has a greater impact on the final distance, that is, the model pays more attention to the feature differences between samples (such as the differences in each band of multispectral data).

[0154] S320: Calculate the weighted comprehensive distance corresponding to each water body monitoring sub - model to obtain the model weighted comprehensive distance.

[0155] S330: Calculate according to the weighted comprehensive distance corresponding to the water body monitoring sub - model and the model weighted comprehensive distance to determine the dynamic weight corresponding to the water body monitoring sub - model.

[0156] Specifically, in step S330, the dynamic weight ω k corresponding to the k - th water body monitoring sub - model can be calculated through the following process:

[0157] First, calculate the weight ω k :

[0158]

[0159] where ω k is the dynamic weight corresponding to the k-th water body monitoring sub-model, is the model weighted comprehensive distance, and D j,k is the weighted comprehensive distance corresponding to the k-th water body monitoring sub-model.

[0160] Secondly, normalize the weight ω k to obtain the dynamic weight ω k ′ corresponding to the k-th water body monitoring sub-model:

[0161]

[0162] where K represents the number of models of the water body monitoring sub-models.

[0163] Optionally, the water quality remote sensing monitoring method provided by the embodiments of the present invention further includes:

[0164] S510: Generate a spatial distribution map of the chlorophyll a concentration of the target water area according to the monitoring results of the chlorophyll a concentration of the water body at each single point corresponding to the data to be measured in the target water area;

[0165] According to the monitoring results of the chlorophyll a concentration of the water body at each single point corresponding to the data to be measured in the target water area, perform visualization processing (for example, represent the size of the chlorophyll a concentration of the water body at this point by the depth of the color at this point / the density of the distribution line, etc.) to generate a spatial distribution map of the chlorophyll a concentration of the target water area.

[0166] S520: Analyze the spatial distribution map of the chlorophyll a concentration of the target water area, and determine the water body monitoring results of the target water area according to the analysis results.

[0167] Based on the water quality remote sensing monitoring method provided by the embodiments of the present invention, using the remote sensing image data of the entire domain and its corresponding spectral indices as inputs, perform the global prediction of the chlorophyll a concentration. This prediction process can accurately invert the chlorophyll a concentration of the entire water body area, thereby generating a spatial distribution map of the chlorophyll a concentration of the target water area. By analyzing the spatial distribution map of the chlorophyll a concentration of the target water area, it is possible to achieve efficient and accurate prediction of the chlorophyll a concentration of the water body in the target area, determine the water body monitoring results of the target water area, and provide a scientific basis for the pollution control of the target water area.

[0168] Optionally, the water quality remote sensing monitoring method provided by the embodiments of the present invention further includes:

[0169] S610: Obtain the measured results of the chlorophyll a concentration of the water body at each single point corresponding to the data to be measured in the target water area;

[0170] S620: Calculate a first evaluation value according to a fourth preset formula based on the average of the monitoring results of the target water area, the monitoring results of the water body chlorophyll a concentration corresponding to each data to be measured in the target water area, and the measured results of the water body chlorophyll a concentration; the average of the monitoring results of the target water area is calculated through the monitoring results of the water body chlorophyll a concentration corresponding to each data to be measured in the target water area.

[0171] Specifically, the first evaluation value is calculated through the calculation formula of the correlation coefficient (R 2 ):

[0172]

[0173] where R 2 is the first evaluation value, y j is the measured result of the water body chlorophyll a concentration corresponding to the data to be measured in the target water area, is the monitoring result of the water body chlorophyll a concentration corresponding to the data to be measured in the target water area, is the average of the monitoring results of the target water area, and M is the number of models of the water body monitoring sub-model.

[0174] S630: Calculate a second evaluation value according to a fifth preset formula based on the monitoring results of the water body chlorophyll a concentration corresponding to each data to be measured in the target water area and the measured results of the water body chlorophyll a concentration;

[0175] Specifically, the second evaluation value is calculated through the calculation formula of the root mean square error (RMSE):

[0176]

[0177] where RMSE is the second evaluation value, y j is the measured result of the water body chlorophyll a concentration corresponding to the data to be measured in the target water area, is the monitoring result of the water body chlorophyll a concentration corresponding to the data to be measured in the target water area, and M is the number of models of the water body monitoring sub-model.

[0178] S640: Analyze based on the first evaluation value and the second value to determine the accuracy of the water body monitoring in the target water area.

[0179] Specifically, the method provided in the embodiment of the present invention evaluates the prediction accuracy by comparing the model prediction results with the measured data and using indicators such as R 2 and RMSE. By visually displaying the chlorophyll a concentration distribution in different regions, the effectiveness and applicability of the model under different environmental conditions are further verified.

[0180] In summary, implementing the embodiments of the present invention includes the following beneficial effects:

[0181] An embodiment of the present invention provides a water quality remote sensing monitoring method, system, device, and storage medium. This method calculates by inputting the data to be measured into a number of water body monitoring sub-models, determines the dynamic weight corresponding to each water body monitoring sub-model according to the obtained sub-model monitoring results and the training sample set, and finally obtains the monitoring result of the chlorophyll a concentration of the water body in the target water area corresponding to the data to be measured through weighted calculation of the sub-model monitoring results; in this regard, the method provided by the present invention determines the dynamic weight of the sub-model and performs weighted summation calculation on the sub-model monitoring results, and determines the final monitoring result according to the calculation result. Such a monitoring process can integrate the algorithm advantages of each water body monitoring sub-model, enhance the generalization ability and stability of the water quality remote sensing monitoring method, and improve the accuracy of monitoring the chlorophyll a concentration of the water body in the target water area.

[0182] As Figure 2 shown, an embodiment of the water quality remote sensing monitoring method provided by the embodiment of the present invention includes steps S10 to S50, which are specifically as follows:

[0183] S10: Obtain multispectral image data

[0184] In this embodiment, the multispectral remote sensing image is collected by the YUSENSE AQ600Pro sensor carried by the DJI Matrice 300RTK drone. During image collection, the flight altitude of the drone is set to 100 meters, and the speed is controlled at 5 m / s to ensure the stability of the collected images and reduce the atmospheric influence to a negligible level. Using this setting, the spatial resolution of the collected images is approximately 4.5 cm. After collection, the multispectral image is preprocessed using Yusense Map software, including inter-channel registration, stitching, and radiometric correction, to ensure the quality and consistency of the data.

[0185] S20: Calculate spectral indices

[0186] Based on the collected multispectral image, calculate 12 spectral indices, such as the two-band index (2BA), blue-green index (BG), red-green index (RG), fluorescence line height blue index (FLH Blue), Be16NDPhyI, normalized difference turbidity index (NDTI), blue-red ratio index (BR), green-red ratio index (GR), ratio vegetation index (RVI), ratio spectral index (RSI), chlorophyll index (GCI), and normalized difference vegetation index (NDVI).

[0187] S30: Predict the chlorophyll a concentration

[0188] S31: Construct the training set

[0189] Using the collected multispectral image data and its derived 12 spectral indices, 70% of the field measured chlorophyll a concentration data was used as a training set. The training set contains remote sensing images of various water body areas, corresponding spectral indices and measured chlorophyll a concentrations, which are input into the EMD method as input data.

[0190] Specifically, the EMD method trains the prediction model of chlorophyll a concentration by integrating multiple machine learning and deep learning algorithms.

[0191] In order to cope with the nonlinear characteristics of data and the complexity of local variation rules in water chlorophyll a concentration monitoring, an embodiment of the present invention provides a water quality remote sensing monitoring method that integrates multiple machine learning and deep learning algorithms (EMD). The method combines multiple water monitoring algorithm sub-models (four water monitoring sub-models are used in the embodiment of the present invention: support vector machine SVM, random forest RF, adaptive boosting algorithm AdaBoost and multi-layer perceptron MLP), and makes full use of the complementary advantages of each algorithm in feature extraction, pattern recognition and generalization ability. Specifically, in the water monitoring sub-model used in the embodiment of the method of the present invention, SVM is good at processing nonlinear data and adapting to small sample learning, RF is known for its robustness to high-dimensional features and noise, AdaBoost optimizes the performance of difficult-to-classify samples through weighted iteration, and MLP can capture the complex nonlinear relationship between high-dimensional features by virtue of its deep structure. In order to achieve efficient collaboration of different algorithms, the EMD method introduces a distance-based weighted fusion strategy, including the processing method of "identifying nearby samples → calculating combined distance → inverse distance weight allocation". This strategy dynamically adjusts the weights of each algorithm by calculating the sample feature distribution and prediction error, thereby enhancing the overall prediction accuracy and robustness of the model. This design effectively integrates the advantages of a single algorithm while overcoming its limitations, and is expected to significantly improve the reliability and practicality of chlorophyll a concentration monitoring.

[0192] S32: Model training and validation

[0193] After the training is completed, the remaining 30% of the measured data is used as the validation set to evaluate the accuracy of the trained EMD model. The validation set contains the measured chlorophyll a concentration data of different water bodies, and corresponds to the corresponding remote sensing images and spectral indexes. By calculating the chlorophyll a concentration inversion accuracy of the training set and the validation set, the generalization ability and stability of the model can be effectively evaluated.

[0194] S40: Global Chlorophyll a Concentration Prediction

[0195] Based on the trained EMD model, the global remote sensing image data and its corresponding spectral indices are used as inputs to perform global prediction of chlorophyll-a concentration. This prediction process can accurately retrieve the chlorophyll-a concentration over the entire water body area, thereby generating a spatial distribution map of the chlorophyll-a concentration in the target water area. Through this method, efficient and accurate prediction of the chlorophyll-a concentration in the water body of the target area can be achieved, providing a scientific basis for water quality monitoring and pollution control.

[0196] S50: Result Evaluation and Visualization

[0197] Compare the model prediction results with the measured data, and use metrics such as R 2 and RMSE to evaluate the prediction accuracy. By visually displaying the chlorophyll-a concentration distribution in different regions, the effectiveness and applicability of the model under different environmental conditions are further verified. In addition, the system also supports real-time update and dynamic monitoring, and can provide continuous data support for water pollution monitoring and ecological environment protection.

[0198] In summary, implementing the embodiments of the present invention includes the following beneficial effects:

[0199] (1) Enhance feature expression ability: By calculating multiple spectral indices and optimizing the optical characteristics of the water body, the model can more accurately retrieve the chlorophyll-a concentration.

[0200] (2) Integrate multiple models to improve prediction ability: Integrate deep learning and machine learning models to improve the generalization ability and reduce the limitations of a single model.

[0201] (3) Dynamic weighted fusion to improve stability: Adopt an error adaptive weighted method to improve the adaptability under different water body environments.

[0202] As Figure 3 shown, the embodiments of the present invention also provide a water quality remote sensing monitoring system, including:

[0203] The first module is used to obtain the data to be measured; the data to be measured is determined according to the water body monitoring remote sensing data;

[0204] The second module is used to input the data to be measured into several water body monitoring sub-models for monitoring, and obtain the sub-model monitoring results corresponding to the data to be measured for each water body monitoring sub-model;

[0205] The third module is used to obtain the training sample set and determine the dynamic weights corresponding to several water body monitoring sub-models according to the training sample set, the data to be measured, the monitoring results of several sub-models, and a preset algorithm;

[0206] A fourth module is configured to determine the monitoring result of the chlorophyll-a concentration of the water body in the target water area corresponding to the data to be measured based on the monitoring results of a number of sub-models and the dynamic weights corresponding to a number of water body monitoring sub-models.

[0207] It can be seen that the content in the above method embodiments is applicable to the system embodiments herein. The functions specifically implemented in the system embodiments are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those in the above method embodiments.

[0208] As Figure 4 shown, an embodiment of the present invention further provides a water quality remote sensing monitoring device, including:

[0209] At least one processor;

[0210] At least one memory for storing at least one program;

[0211] When the at least one program is executed by the at least one processor, the at least one processor implements the method as described above.

[0212] Among them, the memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. The memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory may optionally include a remote memory remotely disposed relative to the processor, and these remote memories may be connected to the processor through a network. Examples of the above networks include, but are not limited to, the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0213] It can be seen that the content in the above method embodiments is applicable to the device embodiments herein. The functions specifically implemented in the device embodiments are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those in the above method embodiments.

[0214] In addition, an embodiment of the present application also discloses a computer program product or a computer program. The computer program product or the computer program is stored in a computer-readable storage medium. The processor of the computer device can read the computer program from the computer-readable storage medium, and the processor executes the computer program, so that the computer device executes the above method.

[0215] An embodiment of the present invention also provides a computer-readable storage medium, which stores a program executable by a processor. The program executable by the processor, when executed by the processor, is used to implement the above-mentioned method. Similarly, the content in the above method embodiments is applicable to this storage medium embodiment. The functions specifically implemented by this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those in the above method embodiments.

[0216] It can be understood that all or some of the steps and systems in the methods disclosed above can be implemented as software, firmware, hardware, and their appropriate combinations. Some physical components or all physical components can be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include a computer storage medium (or non-transitory medium) and a communication medium (or transitory medium). As is well known to those of ordinary skill in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data. Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory, or other memory technologies, CD-ROM, digital versatile disk (DVD), or other optical disk storage, magnetic cassette, tape, magnetic disk storage, or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, as is well known to those of ordinary skill in the art, a communication medium typically contains computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transmission mechanism, and can include any information delivery medium.

[0217] The above is a specific description of the preferred embodiment of the present invention. However, the present invention is not limited to the described embodiment. Those skilled in the art can make various equivalent deformations or substitutions without departing from the spirit of the present invention. These equivalent deformations or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A water quality remote sensing monitoring method, characterized in that: include: Get the data to be tested; The data to be measured is determined according to water quality remote sensing monitoring data; Inputting the data to be measured into a plurality of water body monitoring sub-models for monitoring, and obtaining a sub-model monitoring result corresponding to the data to be measured for each water body monitoring sub-model; Acquire a training sample set; determine the dynamic weights corresponding to the several water body monitoring sub-models according to the training sample set, the data to be tested, the monitoring results of the several sub-models and a preset algorithm; Based on the monitoring results of several sub-models and the dynamic weights corresponding to several water body monitoring sub-models, the monitoring results of the chlorophyll a concentration in the water body of the target water area corresponding to the data to be tested are determined.

2. The method according to claim 1, characterized in that The training sample set includes an input data sample set and a monitoring result sample set; the dynamic weight corresponding to the water body monitoring sub-model is determined by the following method: Calculate based on the data to be measured, the sub-model monitoring results of each of the water body monitoring sub-models corresponding to the data to be measured, the input data sample set and the monitoring result sample set to determine the weighted comprehensive distance corresponding to each of the water body monitoring sub-models; Calculating the weighted comprehensive distance corresponding to each of the water body monitoring sub-models to obtain a model weighted comprehensive distance; The dynamic weight corresponding to the water body monitoring sub-model is determined by performing calculations based on the weighted comprehensive distance corresponding to the water body monitoring sub-model and the weighted comprehensive distance of the model.

3. The method according to claim 2, characterized in that The weighted comprehensive distance corresponding to the water body monitoring sub-model is determined by: Calculating based on the sub-model monitoring result of the water body monitoring sub-model corresponding to the data to be measured and the monitoring result sample set to obtain a prediction error set; Perform judgment based on the prediction error set and preset conditions, and determine the neighborhood sample space corresponding to the data to be tested according to the result of the judgment, the input data sample set and the monitoring result sample set; Calculation is performed based on the data to be measured, the sub-model monitoring results of the water body monitoring sub-model corresponding to the data to be measured, and the neighborhood sample space corresponding to the data to be measured to determine the weighted comprehensive distance corresponding to the water body monitoring sub-model.

4. The method according to claim 3, characterized in that The step of making a judgment based on the prediction error set and the preset conditions, and determining the neighborhood sample space corresponding to the data to be tested according to the result of the judgment, the input data sample set and the monitoring result sample set, comprises: Performing weighted sorting on the error values ​​in the prediction error set to obtain a weighted sorting result; A judgment is made according to the weighted sorting result and preset conditions, and according to the result of the judgment, a neighborhood sample space corresponding to the data to be tested is constructed based on a number of input data samples that meet the conditions and their corresponding monitoring result samples.

5. The method according to claim 3, characterized in that Determining the weighted comprehensive distance corresponding to the water body monitoring sub-model includes: Calculating based on a first preset formula, the data to be measured and a neighborhood sample space corresponding to the data to be measured, to obtain a feature weighted Euclidean distance corresponding to the water body monitoring sub-model; Based on the second preset formula, the water body monitoring prediction result of the water body monitoring sub-model corresponding to the data to be tested and the neighborhood sample space corresponding to the data to be tested, a label difference metric corresponding to the water body monitoring sub-model is calculated; Based on the third preset formula, preset coefficients, and the feature weighted Euclidean distance and label difference measurement corresponding to the water body monitoring sub-model, a weighted comprehensive distance corresponding to the water body monitoring sub-model is determined.

6. The method according to claim 1, characterized in that The method further comprises: Generate a spatial distribution map of chlorophyll a concentration in the target water area according to the chlorophyll a concentration monitoring result of each of the measured data corresponding to the target water area; The spatial distribution map of chlorophyll a concentration in the target water area is analyzed, and the water body monitoring result of the target water area is determined according to the analysis result.

7. The method according to claim 1, characterized in that The method further comprises: Obtain the actual measured result of the chlorophyll a concentration in the water body corresponding to each target water area of ​​the data to be measured; According to the fourth preset formula, a first evaluation value is obtained by calculating the mean value of the monitoring results of the target water area and the monitoring results of the chlorophyll a concentration in the water body of each of the data to be tested corresponding to the target water area and the actual measured results of the chlorophyll a concentration in the water body; the mean value of the monitoring results of the target water area is calculated by calculating the monitoring results of the chlorophyll a concentration in the water body of each of the data to be tested corresponding to the target water area; According to the fifth preset formula, a second evaluation value is obtained by calculating the chlorophyll a concentration monitoring result of the target water area corresponding to each of the measured data and the measured chlorophyll a concentration of the target water area; An analysis is performed based on the first evaluation value and the second evaluation value to determine the water body monitoring accuracy of the target water area.

8. A water quality remote sensing monitoring system, characterized in that: include: The first module is used to obtain the data to be tested; The data to be measured is determined according to water quality remote sensing monitoring data; The second module is used to input the data to be measured into several water body monitoring sub-models for monitoring, and obtain the sub-model monitoring results of each water body monitoring sub-model corresponding to the data to be measured; The third module is used to obtain a training sample set and determine the dynamic weights corresponding to the several water body monitoring sub-models according to the training sample set, the data to be tested, the monitoring results of the several sub-models and a preset algorithm; The fourth module is used to determine the chlorophyll a concentration monitoring result of the water body corresponding to the target water area of ​​the data to be tested based on the monitoring results of several sub-models and the dynamic weights corresponding to several water body monitoring sub-models.

9. A water quality remote sensing monitoring device, characterized in that: include: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a program executable by a processor, characterized in that: The processor-executable program is used to perform the method according to any one of claims 1 to 7 when executed by the processor.

Images