An online water quality analysis method and system based on a multispectral image analysis model
By constructing a multispectral image analysis model, efficient inversion of plankton coverage, transparency, and eutrophication in water bodies was achieved, overcoming the limitations of existing water quality monitoring technologies and providing high-precision water quality analysis results.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- FRESHWATER FISHERIES RES CENT OF CHINESE ACAD OF FISHERY SCI
- Filing Date
- 2024-12-10
- Publication Date
- 2026-05-12
AI Technical Summary
Existing multispectral image analysis technology is difficult to address different types of water quality problems simultaneously in water quality analysis, such as phytoplankton coverage, water transparency inversion, and eutrophication. Furthermore, traditional water quality monitoring methods cannot quickly and comprehensively reflect the overall condition of water bodies.
A multispectral image analysis model was adopted, including multispectral image preprocessing, feature extraction, classification model and multiple inversion models, to construct inversion models of plankton coverage, transparency and eutrophication in water bodies, and detailed water quality analysis results were obtained through fusion processing.
It enables high-precision, real-time monitoring of water quality, provides detailed water quality data, simplifies data processing procedures, and improves analysis efficiency and accuracy.
Smart Images
Figure CN119555679B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of water quality analysis technology, and more specifically to an online water quality analysis method and system based on a multispectral image analysis model. Background Technology
[0002] Rapid industrialization and urbanization have led to the deterioration of surface water quality in inland rivers, lakes, and reservoirs, resulting in large-scale outbreaks of environmental problems such as cyanobacterial blooms and eutrophication. These issues directly impact aquatic ecological security, drinking water safety, and socio-economic development. Water resource and environmental monitoring is a crucial prerequisite for the planning, management, and protection of water bodies and aquatic biological resources. Against the backdrop of global warming and intensified human interference, rapid, efficient, and real-time monitoring of water resources and water quality is of great significance for water resource assessment, water pollution control, and water security.
[0003] Traditional water quality monitoring involves collecting samples from the field and then analyzing them in a laboratory to determine their physicochemical properties. While this method offers high accuracy, it is time-consuming and labor-intensive. Furthermore, the measured data can only represent information from local sample points and cannot quickly reflect the overall spatial water quality status of the water body, thus failing to meet the requirements for real-time dynamic water quality monitoring.
[0004] In recent years, the development of UAV remote sensing technology has provided new possibilities for water quality monitoring. Multispectral imaging technology, as an advanced remote sensing method, can capture the spectral characteristics of water bodies in different bands. These characteristics are closely related to various chemical and physical parameters in the water. By analyzing this spectral data, the water quality status of the water body can be inferred.
[0005] However, current multispectral image analysis techniques have some limitations in water quality analysis. First, a single spectral analysis model cannot simultaneously address different types of water quality issues, such as phytoplankton coverage, water transparency inversion, and eutrophication. Second, existing water quality monitoring methods often only obtain limited water quality parameters, making it difficult to comprehensively reflect the complex conditions of water bodies.
[0006] Therefore, how to provide an online water quality analysis method that can comprehensively reflect the complex conditions of water bodies is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention
[0007] In view of this, the present invention provides an online water quality analysis method and system based on a multispectral image analysis model, which realizes real-time and accurate analysis of water quality by acquiring and analyzing multispectral images of the target water body.
[0008] To achieve the above objectives, the present invention adopts the following technical solution:
[0009] An online water quality analysis method based on a multispectral image analysis model includes:
[0010] Acquire multispectral images of the target water body and preprocess the multispectral images;
[0011] A multispectral image analysis model is constructed, which includes: an analysis model, a classification model, a water body plankton coverage inversion model, a water body transparency inversion model, a water surface eutrophication inversion model, and a water quality parameter analysis model.
[0012] The analytical model is used to extract features from the preprocessed multispectral image to obtain the current water quality data and water body data of the target water body;
[0013] The water body data is input into a classification model to extract the water surface line of the target water body;
[0014] The current water quality data and the water surface line are input into the water body plankton coverage inversion model, the water body transparency inversion model, and the water surface eutrophication inversion model. Water body plankton coverage inversion, water body transparency inversion, and water surface eutrophication inversion are performed respectively to obtain water body plankton coverage inversion results, water body transparency inversion results, and water surface eutrophication inversion results.
[0015] The water quality parameter analysis model is used to fuse the inversion results of the water body plankton coverage, the water body transparency, and the water surface eutrophication to output the final water quality analysis results.
[0016] Preferably, the preprocessing of the multispectral image includes: geometric correction and radiometric correction, multispectral image fusion, cloud and shadow removal, noise filtering, and image enhancement.
[0017] Preferably, multispectral image fusion includes:
[0018] Perform image registration on the first image and the second image respectively;
[0019] Feature extraction is performed on the first and second images after image registration to obtain the first feature;
[0020] The first feature is used to classify the data, and the classification result is obtained.
[0021] Decision-level fusion is performed based on the classification results to obtain the fused multispectral image.
[0022] Preferably, the analytical model is used to extract features from the preprocessed multispectral image to obtain the current water quality data and water body data of the target water body, including:
[0023] The preprocessed multispectral image is input into the first feature extraction module to extract local spatial information and global spectral information, thereby obtaining local spatial features and global spectral features.
[0024] The multispectral image is input into the second feature extraction module to extract local spectral information and global spatial information, thereby obtaining the features of the central region and the features of the adjacent regions.
[0025] The local spatial features, the global spectral features, the central region features, and the neighboring region features are input into the prediction and classification module to classify the multispectral image and output the class label of the multispectral image.
[0026] The current water quality data and water body data of the target water body are obtained based on the class labels.
[0027] Preferably, the water body data is input into a classification model to extract the water surface line of the target water body, including:
[0028] Feature extraction is performed on the water body data to obtain water body raster features;
[0029] Feature extraction is performed on the water body raster features to obtain at least two local water body raster features;
[0030] The features of at least two local water body raster features are fused to obtain global water body raster features;
[0031] Based on the classification and regression processing of the global water body raster features, the key points of the water surface line are obtained;
[0032] By fitting the key points, the water surface system of the target water body is obtained.
[0033] Preferably, the current water quality data and the water surface line are input into the water body plankton coverage inversion model to perform water body plankton coverage inversion, including:
[0034] Acquire historical water quality monitoring data for the target water body over multiple historical time periods;
[0035] Generate an independent variable dataset and a dependent variable dataset based on the historical water quality monitoring data;
[0036] Extract the first independent variable component from the independent variable dataset, and extract the first dependent variable component from the dependent variable dataset;
[0037] The partial least squares regression algorithm is used to perform autoregression processing on the first autovariable and the first dependent variable, and the model accuracy is determined based on the autoregression results;
[0038] If the model accuracy does not meet the preset accuracy condition, then the second independent variable component is extracted from the independent variable dataset, and the second dependent variable component is extracted from the dependent variable dataset for iterative processing until the model meets the preset stopping iteration condition, thus obtaining the water body planktonic coverage inversion model.
[0039] The current water quality data and the water surface line are input into the water body plankton coverage inversion model to perform water body plankton coverage inversion and obtain the water body plankton coverage inversion result of the target water body.
[0040] Preferably, the current water quality data and the water surface line are input into the eutrophication inversion model to perform eutrophication inversion, including:
[0041] Multiple eutrophication inversion models were constructed based on various machine learning algorithms. The entropy weight method was used to organically combine the multiple eutrophication inversion models to obtain the optimal eutrophication inversion model.
[0042] The current water quality data and the water surface line are input into the optimal water surface eutrophication inversion model to perform water surface eutrophication inversion and obtain the current water surface eutrophication inversion result of the target water body.
[0043] Preferably, the water quality parameter analysis model is used to fuse the inversion results of the water body plankton coverage, the water body transparency, and the water surface eutrophication to output the final water quality analysis results, including:
[0044] Correlation analysis was performed on the inversion results of phytoplankton coverage, water transparency, and eutrophication, respectively, and the correlation analysis results were obtained.
[0045] The parameters of the water quality analysis model were adjusted based on the correlation analysis results.
[0046] The results of the inversion of phytoplankton coverage, water transparency, and eutrophication are input into the adjusted water quality analysis model to obtain the final water quality analysis results.
[0047] On the other hand, this invention discloses an online water quality analysis system based on a multispectral image analysis model, used to implement the aforementioned online water quality analysis method based on a multispectral image analysis model, comprising:
[0048] The acquisition module is used to acquire multispectral images of the target water body;
[0049] A transfer station is used for preprocessing the multispectral images;
[0050] The analysis module is used for water quality analysis based on preprocessed multispectral images; it includes:
[0051] The analytical unit is used to extract features from the preprocessed multispectral image using an analytical model to obtain the current water quality data and water body data of the target water body.
[0052] The water surface line extraction unit is used to input the water body data into the classification model and extract the water surface line of the target water body.
[0053] The inversion unit is used to input the current water quality data and the water surface line into the water body plankton coverage inversion model, the water body transparency inversion model, and the water surface eutrophication inversion model, respectively, to perform water body plankton coverage inversion, water body transparency inversion, and water surface eutrophication inversion, and obtain water body plankton coverage inversion results, water body transparency inversion results, and water surface eutrophication inversion results;
[0054] The water quality analysis unit is used to integrate the inversion results of the water body plankton coverage, the water body transparency, and the water surface eutrophication using the water quality parameter analysis model, and output the final water quality analysis results.
[0055] As can be seen from the above technical solution, compared with the prior art, this invention discloses an online water quality analysis method and system based on a multispectral image analysis model. Through the multispectral image analysis model, high-precision monitoring of water quality is achieved, providing detailed water quality data. Automated multispectral image preprocessing and feature extraction simplify the data processing flow and improve analysis efficiency. By integrating water body plankton coverage inversion, water body transparency inversion, and water surface eutrophication inversion, the mutual influence between the results of these inversions is considered, making water quality analysis more accurate. Attached Figure Description
[0056] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0057] Figure 1 This is a flowchart illustrating the method provided by the present invention.
[0058] Figure 2 The system framework diagram provided for this invention. Detailed Implementation
[0059] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0060] This invention discloses an online water quality analysis method based on a multispectral image analysis model, such as... Figure 1 As shown, it includes:
[0061] Multispectral images of the target water body were acquired and preprocessed. A drone equipped with a multispectral camera was used to conduct aerial photography of the target water body, obtaining high-quality multispectral images. This camera can capture spectral information in different bands, providing a rich data source for water quality analysis. The acquired images were transmitted in real-time to a relay workstation via a wireless network. At the relay workstation, the received multispectral images underwent rapid preprocessing and stitching. Preprocessing steps included geometric correction, radiometric correction, cloud and shadow removal, noise filtering, and image enhancement to ensure the accuracy and usability of the image data.
[0062] A multispectral image analysis model was constructed, which includes: an analytical model, a classification model, a water body plankton coverage inversion model, a water body transparency inversion model, a water surface eutrophication inversion model, and a water quality parameter analysis model.
[0063] The analytical model is used to extract features from the preprocessed multispectral image to obtain the current water quality data and water body data of the target water body.
[0064] Input water body data into a classification model to extract the water surface line of the target water body;
[0065] The current water quality data and water surface line are input into the water body plankton coverage inversion model, water body transparency inversion model and water surface eutrophication inversion model. The water body plankton coverage inversion, water body transparency inversion and water surface eutrophication inversion are performed respectively to obtain the water body plankton coverage inversion results, water body transparency inversion results and water surface eutrophication inversion results.
[0066] The results of phytoplankton coverage inversion, water transparency inversion, and eutrophication inversion are fused using a water quality parameter analysis model to output the final water quality analysis results. These results include water temperature, water depth, algae coverage, water transparency, total phosphorus, and total nitrogen.
[0067] Furthermore, preprocessing of multispectral images includes: geometric and radiometric correction, multispectral image fusion, cloud and shadow removal, noise filtering, and image enhancement.
[0068] Geometric correction is used to correct geometric distortions in an image, such as tilt, warping, or misalignment, ensuring the accuracy of spatial information within the image. Radiometric correction, on the other hand, focuses on adjusting the brightness, contrast, and color balance of an image to eliminate radiometric distortions caused by factors such as sensor characteristics, atmospheric conditions, or the angle of the sun, making the image closer to the real scene.
[0069] Multispectral image fusion involves integrating image information from different spectral bands (such as visible light, infrared, near-infrared, and shortwave infrared) to generate a single image containing richer information and easier to interpret. This ensures effective fusion of information from different bands while avoiding information redundancy and distortion, thereby improving the overall information content and readability of the image. Multiple images are then stitched together to obtain a comprehensive multispectral image of the target water body.
[0070] In multispectral images, clouds and shadows can obscure target areas, affecting image quality and subsequent analysis. Advanced algorithms can identify and remove clouds and shadows, restoring information from obscured areas or at least reducing their interference with overall image analysis.
[0071] Images often contain various types of noise due to sensor noise, environmental factors, or errors during data transmission. Noise filtering techniques aim to remove these unwanted random variations, preserve useful information in the image, and improve image sharpness and signal-to-noise ratio.
[0072] To further enhance the visualization and readability of images, image enhancement processing is typically performed. This includes contrast enhancement, sharpening, and color enhancement, aiming to make features in the image more prominent, edges clearer, and easier for human observation or automatic machine recognition.
[0073] Furthermore, multispectral image fusion includes:
[0074] Image registration is performed on the first image and the second image respectively, including:
[0075] First, the first and second images (which come from different spectral bands) are preprocessed, including noise removal, image size and resolution adjustment, etc., to ensure that they have similar image quality and spatial resolution when they are registered.
[0076] Key feature points are detected in the first and second images using feature point detection algorithms (such as SIFT, SURF, etc.). These feature points are usually significant structural or texture changes in the images.
[0077] By comparing the descriptors of the feature points, the feature points in the first image are matched with the feature points in the second image to establish a correspondence between the two images.
[0078] Based on the matched feature point pairs, calculate the transformation matrix between the images (such as affine transformation, perspective transformation, etc.), and transform the second image to make it completely aligned with the first image in space.
[0079] Feature extraction is performed on the first and second images after image registration to obtain the first feature. Multi-scale analysis is then performed on the first and second images after image registration to extract image features at different scales, such as edges, textures, and shapes. Features from different spectral bands are fused to obtain a comprehensive feature set containing multispectral information, i.e., the first feature.
[0080] The first feature is classified to obtain the classification result. Based on the classification result, a decision-level fusion strategy is designed, such as voting mechanism, weighted average, Bayesian fusion, etc., to comprehensively consider the information from different spectral bands. The fusion strategy is applied to fuse the classification results from different spectral bands to obtain the final fused multispectral image.
[0081] Decision-level fusion is performed based on the classification results to obtain the fused multispectral image.
[0082] Furthermore, an analytical model is used to extract features from the preprocessed multispectral image to obtain the current water quality data and water body data of the target water body, including:
[0083] The preprocessed multispectral image is input into the first feature extraction module to extract local spatial information and global spectral information, resulting in local spatial features and global spectral features. This first feature extraction module is specifically designed to extract local spatial information and global spectral information from the image. Local spatial information helps identify small-scale features in the image, while global spectral information provides an overview of the spectral features of the entire image. Through this module, both local spatial features and global spectral features can be obtained.
[0084] The multispectral image is input into the second feature extraction module to extract local spectral information and global spatial information, thereby obtaining the features of the central region and the features of the adjacent regions.
[0085] Local spatial features, global spectral features, central region features, and neighboring region features are input into the prediction and classification module to classify the multispectral image and output class labels for the multispectral image. Local spatial features, global spectral features, central region features, and neighboring region features obtained from the two feature extraction modules are also input into the prediction and classification module. This module uses these features to classify the multispectral image and outputs class labels for the multispectral image. Class labels are high-level descriptions of the image content, helping to identify different types of water body features in the image.
[0086] The current water quality data and water body data of the target water body are obtained based on the class label, including the spectral characteristics and spatial characteristics of the water body, as well as water quality parameters derived from these characteristics, such as chlorophyll concentration and turbidity.
[0087] In another embodiment, water body data is input into a classification model to extract the water surface line of the target water body, including:
[0088] Feature extraction is performed on water body data to obtain water body raster features;
[0089] Feature extraction is performed on the water body raster features to obtain at least two local water body raster features;
[0090] At least two local water body raster features are fused to obtain global water body raster features;
[0091] Based on the classification and regression processing of global water body raster features, key points of the water surface line are obtained;
[0092] By fitting the key points, the water surface system of the target water body is obtained.
[0093] In another embodiment, the current water quality data and water surface line are input into the water body plankton coverage inversion model to perform water body plankton coverage inversion, including:
[0094] Acquire historical water quality monitoring data for the target water body over multiple historical time periods, including but not limited to key water quality parameters such as dissolved oxygen, ammonia nitrogen, transparency, and the types and quantities of plankton, as well as data collection time and spatial location information.
[0095] The independent variable dataset and the dependent variable dataset are generated based on historical water quality monitoring data. The independent variable dataset contains various factors that affect water quality, while the dependent variable dataset contains the corresponding water quality parameter values.
[0096] Extract the first independent variable component from the independent variable dataset and extract the first dependent variable component from the dependent variable dataset;
[0097] Partial least squares regression algorithm is used to perform autoregression processing on the first autovariable and the first dependent variable, and the model accuracy is determined based on the autoregression results;
[0098] If the model accuracy does not meet the preset accuracy condition, the second independent variable component is extracted from the independent variable dataset, and the second dependent variable component is extracted from the dependent variable dataset for iterative processing until the model meets the preset stopping iteration condition, thus creating the water body planktonic coverage inversion model.
[0099] The current water quality data and water surface line are input into the water body plankton coverage inversion model to perform water body plankton coverage inversion and obtain the water body plankton coverage inversion results for the target water body.
[0100] In another embodiment, the water transparency inversion process is similar to the water plankton coverage inversion process described above in practical applications.
[0101] In another embodiment, the current water quality data and water surface line are input into the eutrophication inversion model to perform eutrophication inversion, including:
[0102] Multiple eutrophication inversion models were constructed using various machine learning algorithms (such as support vector machines, random forests, gradient boosting trees, and neural networks). These models were then organically combined using the entropy weighting method to obtain the optimal eutrophication inversion model. To obtain this optimal model, the entropy weighting method was used to evaluate the predictive performance of each model and assign weights accordingly. The entropy weighting method reflects the stability and reliability of a model's prediction by calculating the entropy value (i.e., the degree of uncertainty) of its prediction results. Weight allocation follows the principle of "the smaller the entropy value, the larger the weight," meaning that the more stable and reliable the prediction results, the greater the contribution of the model in the final combined model. Based on the calculated weights, the multiple eutrophication inversion models were organically combined to obtain the optimal model. This combined model integrates the advantages of each model, improving the accuracy and stability of predictions.
[0103] The current water quality data and water surface line are input into the optimal water surface eutrophication inversion model to perform water surface eutrophication inversion and obtain the current water surface eutrophication inversion result of the target water body.
[0104] In another embodiment, the results of phytoplankton coverage inversion, water transparency inversion, and eutrophication inversion are fused using a water quality parameter analysis model to output the final water quality analysis results, including:
[0105] The inversion results of phytoplankton coverage, water transparency, and eutrophication were analyzed separately. Specifically, correlation analysis methods in statistics (such as Pearson correlation coefficient and Spearman rank correlation coefficient) were used to quantify the degree of correlation among phytoplankton coverage, water transparency, and eutrophication status.
[0106] The parameters of the water quality analysis model were adjusted based on the correlation analysis results.
[0107] The results of inversion of phytoplankton coverage, water transparency, and eutrophication were input into the adjusted water quality analysis model to obtain the final water quality analysis results.
[0108] On the other hand, this invention discloses an online water quality analysis system based on a multispectral image analysis model, used to implement the aforementioned online water quality analysis method based on a multispectral image analysis model, such as... Figure 2 As shown, it includes:
[0109] The acquisition module is used to acquire multispectral images of the target water body;
[0110] A transfer station used for preprocessing multispectral images;
[0111] The analysis module is used for water quality analysis based on preprocessed multispectral images; it includes:
[0112] The analytical unit is used to extract features from the preprocessed multispectral image using an analytical model to obtain the current water quality data and water body data of the target water body.
[0113] The water surface line extraction unit is used to input water body data into the classification model and extract the water surface line of the target water body;
[0114] The inversion unit is used to input the current water quality data and water surface line into the water body plankton coverage inversion model, water body transparency inversion model, and water surface eutrophication inversion model, respectively, to perform water body plankton coverage inversion, water body transparency inversion, and water surface eutrophication inversion, and obtain water body plankton coverage inversion results, water body transparency inversion results, and water surface eutrophication inversion results;
[0115] The water quality analysis unit is used to integrate the inversion results of phytoplankton coverage, water transparency, and eutrophication using a water quality parameter analysis model, and output the final water quality analysis results.
[0116] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.
[0117] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. An online water quality analysis method based on a multispectral image analysis model, characterized in that, include: Acquire multispectral images of the target water body and preprocess the multispectral images; A multispectral image analysis model is constructed, which includes: an analysis model, a classification model, a water body plankton coverage inversion model, a water body transparency inversion model, a water surface eutrophication inversion model, and a water quality parameter analysis model. The analytical model is used to extract features from the preprocessed multispectral image to obtain the current water quality data and water body data of the target water body; The water body data is input into a classification model to extract the water surface line of the target water body; The current water quality data and the water surface line are input into the water body plankton coverage inversion model, the water body transparency inversion model, and the water surface eutrophication inversion model. Water body plankton coverage inversion, water body transparency inversion, and water surface eutrophication inversion are performed respectively to obtain water body plankton coverage inversion results, water body transparency inversion results, and water surface eutrophication inversion results. The water quality parameter analysis model is used to fuse the inversion results of the water body plankton coverage, the water body transparency, and the water surface eutrophication to output the final water quality analysis results. The analytical model is used to extract features from the preprocessed multispectral image to obtain the current water quality data and water body data of the target water body, including: The preprocessed multispectral image is input into the first feature extraction module to extract local spatial information and global spectral information, thereby obtaining local spatial features and global spectral features. The multispectral image is input into the second feature extraction module to extract local spectral information and global spatial information, thereby obtaining the features of the central region and the features of the adjacent regions. The local spatial features, the global spectral features, the central region features, and the neighboring region features are input into the prediction and classification module to classify the multispectral image and output the class label of the multispectral image. Obtain the current water quality data and water body data of the target water body based on the aforementioned category labels; The water body data is input into a classification model to extract the water surface line of the target water body, including: Feature extraction is performed on the water body data to obtain water body raster features; Feature extraction is performed on the water body raster features to obtain at least two local water body raster features; The features of at least two local water body raster features are fused to obtain global water body raster features; Based on the classification and regression processing of the global water body raster features, the key points of the water surface line are obtained; By fitting the key points, the water surface system of the target water body is obtained; The current water quality data and the water surface line are input into the water body plankton coverage inversion model to perform water body plankton coverage inversion, including: Acquire historical water quality monitoring data for the target water body over multiple historical time periods; Generate independent and dependent variable datasets based on historical water quality monitoring data; Extract the first independent variable component from the independent variable dataset, and extract the first dependent variable component from the dependent variable dataset; The partial least squares regression algorithm is used to perform autoregression processing on the first autovariable and the first dependent variable, and the model accuracy is determined based on the autoregression results; If the model accuracy does not meet the preset accuracy condition, then the second independent variable component is extracted from the independent variable dataset, and the second dependent variable component is extracted from the dependent variable dataset for iterative processing until the model meets the preset stopping iteration condition, thus obtaining the water body planktonic coverage inversion model. The current water quality data and the water surface line are input into the water body plankton coverage inversion model to perform water body plankton coverage inversion and obtain the water body plankton coverage inversion result of the target water body. The current water quality data and the water surface line are input into the eutrophication inversion model to perform eutrophication inversion, including: Multiple eutrophication inversion models were constructed based on various machine learning algorithms. The entropy weight method was used to organically combine the multiple eutrophication inversion models to obtain the optimal eutrophication inversion model. The current water quality data and the water surface line are input into the optimal water surface eutrophication inversion model to perform water surface eutrophication inversion and obtain the current water surface eutrophication inversion result of the target water body. The water quality parameter analysis model is used to fuse the inversion results of the phytoplankton coverage, the water transparency, and the eutrophication of the water surface, outputting the final water quality analysis results, including: Correlation analysis was performed on the inversion results of phytoplankton coverage, water transparency, and eutrophication, respectively, and the correlation analysis results were obtained. The parameters of the water quality analysis model were adjusted based on the correlation analysis results. The results of the inversion of phytoplankton coverage, water transparency, and eutrophication are input into the adjusted water quality analysis model to obtain the final water quality analysis results.
2. The online water quality analysis method based on a multispectral image analysis model according to claim 1, characterized in that, Preprocessing of the multispectral image includes: geometric correction and radiometric correction, multispectral image fusion, cloud and shadow removal, noise filtering, and image enhancement.
3. The online water quality analysis method based on a multispectral image analysis model according to claim 2, characterized in that, Multispectral image fusion, including: Perform image registration on the first image and the second image respectively; Feature extraction is performed on the first and second images after image registration to obtain the first feature; The first feature is used to classify the data, and the classification result is obtained. Decision-level fusion is performed based on the classification results to obtain the fused multispectral image.
4. An online water quality analysis system based on a multispectral image analysis model, characterized in that, A method for implementing an online water quality analysis based on a multispectral image analysis model as described in any one of claims 1-3 includes: The acquisition module is used to acquire multispectral images of the target water body; A transfer station is used for preprocessing the multispectral images; The analysis module is used for water quality analysis based on preprocessed multispectral images; it includes: The analytical unit is used to extract features from the preprocessed multispectral image using an analytical model to obtain the current water quality data and water body data of the target water body. The water surface line extraction unit is used to input the water body data into the classification model and extract the water surface line of the target water body. The inversion unit is used to input the current water quality data and the water surface line into the water body plankton coverage inversion model, the water body transparency inversion model, and the water surface eutrophication inversion model, respectively, to perform water body plankton coverage inversion, water body transparency inversion, and water surface eutrophication inversion, and obtain water body plankton coverage inversion results, water body transparency inversion results, and water surface eutrophication inversion results; The water quality analysis unit is used to integrate the inversion results of the water body plankton coverage, the water body transparency, and the water surface eutrophication using the water quality parameter analysis model, and output the final water quality analysis results.