A water quality monitoring method and system based on remote sensing data

Abstract

The present invention discloses a water quality monitoring method and system based on remote sensing data, which relates to the technical field of water quality monitoring, including collecting hyperspectral data and multispectral data of a monitoring area based on a remote sensing platform, respectively forming a data set with the collected hyperspectral data and multispectral data, and performing data correction; calculating the probability distribution of the corrected hyperspectral data set and multispectral data set for registration, and performing data fusion. The method of the present invention realizes the registration of hyperspectral data and multispectral data by calculating the probability distribution, ensuring that they can be compared and analyzed in the same spatial reference system, and makes the data consistent in space through data fusion, better reflects the actual condition of the water body, and achieves the effect of accurate classification by performing comprehensive feature extraction on the data, cooperating with the construction of a deep learning model, classifying and identifying pollutants of different categories, and statistically analyzing the classification results of various pollutants in each monitoring sample.

Description

Technical Field

[0001] The present invention relates to the technical field of water quality monitoring, and in particular to a water quality monitoring method and system based on remote sensing data. Background Art

[0002] With the gradual enhancement of environmental protection awareness, water quality monitoring has become an important part of environmental protection. Traditional water quality monitoring methods mainly rely on on-site sampling and laboratory analysis. Although this method can provide accurate water quality data, it has many shortcomings such as cumbersome operation, poor real-time performance, and limited spatial coverage. In order to solve these problems, modern science and technology has gradually introduced remote sensing technology to monitor water bodies through hyperspectral and multispectral data. The development of these technologies has made large-scale, real-time, and continuous water quality monitoring possible;

[0003] Hyperspectral remote sensing technology can accurately identify and distinguish the spectral information of different substances by carefully analyzing the spectral characteristics of the target object; multispectral remote sensing can quickly obtain the overall spectral characteristics of the target area by capturing information from a few wide spectral bands. However, single hyperspectral or multispectral data increases the complexity of data processing in practical applications, increases the difficulty of data fusion and feature extraction, and makes the feature extraction effect after fusion poor. When faced with complex water quality environments, it is difficult to provide sufficient classification accuracy and stability, resulting in unreliable water quality assessment results. Summary of the invention

[0004] In view of the problems existing in the above-mentioned existing water quality monitoring methods and systems based on remote sensing data, the present invention is proposed.

[0005] Therefore, the problem to be solved by the present invention is that single hyperspectral or multispectral data increases the complexity of data processing in practical applications, increases the difficulty of data fusion and feature extraction, and makes the feature extraction effect after fusion poor. When faced with complex water quality environments, it is difficult to provide sufficient classification accuracy and stability, resulting in unreliable water quality assessment results.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a water quality monitoring method based on remote sensing data, comprising:

[0007] Based on the remote sensing platform, the hyperspectral data and multispectral data of the monitoring area are collected, and the collected hyperspectral data and multispectral data are respectively formed into data sets, and data correction is performed;

[0008] The probability distribution of the corrected hyperspectral dataset and multispectral dataset is calculated and registered, and data fusion is performed, and fusion features are extracted based on principal component features and spectral features;

[0009] Build a deep learning model to classify data based on data fusion features and evaluate the water quality in the monitoring area;

[0010] Encrypt the classification data and feature data and upload them to the cloud server to generate reports.

[0011] As a preferred solution of the water quality monitoring method based on remote sensing data of the present invention, wherein: the hyperspectral data and multispectral data of the monitoring area collected based on the remote sensing platform include:

[0012] The remote sensing platform is used to collect data from monitoring areas in different regions. The remote sensing platform is equipped with hyperspectral sensors and multispectral sensors. The hyperspectral sensors are used to obtain hyperspectral data of the monitoring areas, and the multispectral sensors are used to obtain multispectral data of the same area.

[0013] Collect solar irradiance data, reflectivity data and noise data for the monitoring area, and calculate the distance data between the sun and the monitoring area;

[0014] Download data from the sensor to the ground control station and back up the original data to an external storage device.

[0015] As a preferred solution of the water quality monitoring method based on remote sensing data of the present invention, wherein: the collected hyperspectral data and multispectral data are respectively formed into data sets, and data correction is performed, including:

[0016] The collected hyperspectral data constitute a hyperspectral dataset ;

[0017] Use geographic reference points to perform geometric correction and convert the corrected hyperspectral data Project it to the standard geographic coordinate system to obtain the preprocessed hyperspectral dataset E;

[0018] The collected multispectral data form a multispectral dataset, which is expressed as , where M and N are spatial dimension data, K is the number of multispectral bands, and the multispectral dataset is reprojected to the standard geographic coordinate system using geographic reference points for geometric correction to obtain the preprocessed multispectral dataset ;

[0019] The atmospheric transmittance and atmospheric reflectance data of the monitoring area are obtained, and atmospheric correction is performed on the preprocessed multispectral dataset.

[0020] The calculation of probability distribution of the corrected hyperspectral data set and multispectral data set is performed to register, and data fusion is performed, including:

[0021] Create a two-dimensional matrix;

[0022] Calculate marginal probability distributions for hyperspectral and multispectral data;

[0023] Hyperspectral data and multispectral data are registered by maximizing mutual information;

[0024] Define the spatial transformation parameters including translation, rotation and scaling, and set the parameters to ,in and represents the amount of translation, represents the rotation angle, and is the scaling factor;

[0025] Initialization setting parameters , which means no translation, no rotation and no scaling, and is applied to the hyperspectral data to get the initialized hyperspectral data ;

[0026] Set the optimization goal to maximize mutual information,

[0027] The gradient descent method is used to perform iterative optimization calculations, and the minimum value is set as the target threshold based on historical data. The iteration is stopped when the iterative change is less than the target threshold, and the optimal transformation parameters are set. Applied to the hyperspectral dataset E, the spatial alignment of the hyperspectral dataset E and the multispectral dataset I is completed;

[0028] Perform wavelet transform on the hyperspectral dataset E and multispectral dataset I respectively to obtain the wavelet coefficients and , and perform data fusion;

[0029] The fused data is represented as a three-dimensional matrix F.

[0030] The fusion feature extraction based on the principal component feature and the spectral feature includes:

[0031] Create an initialized two-dimensional matrix X and convert the three-dimensional matrix Expand and fill the two-dimensional matrix X, the matrix size is expressed as ;

[0032] Remove the mean of the data of each band of the two-dimensional matrix X to form a data matrix after removing the mean , and calculate the mean-removed data matrix The covariance matrix C of , performs eigenvalue decomposition on the covariance matrix C;

[0033] Set the contribution threshold based on historical data, calculate the cumulative variance contribution rate of the selected eigenvalue number d, and determine the number of d when the contribution rate reaches the contribution threshold, and form the eigenvector matrix with the eigenvector corresponding to the selected eigenvalue , the data matrix after removing the mean Projection to the eigenvector matrix , we get the principal component feature matrix Z after dimensionality reduction, and the matrix size is expressed as ;

[0034] Normalize each band from the fused three-dimensional matrix F and calculate the reflectance;

[0035] Based on the normalized reflectance matrix T, the point with the lowest reflectance in the spectrum curve is determined as the absorption peak position;

[0036] Based on the normalized reflectance matrix T, for each pair of bands and , calculate the spectral slope;

[0037] based on Composing the spectral slope matrix , where q is the number of band logarithms;

[0038] The spectral feature matrix is ​​formed based on the normalized reflectance matrix T, the absorption peak position and the spectral slope matrix S , the matrix size is .

[0039] The deep learning model is constructed to classify data based on data fusion features, and the water quality of the monitoring area is evaluated, including:

[0040] The principal component feature matrix Z and the spectral feature matrix after dimension reduction are Combine into a complete feature matrix X;

[0041] Build a deep learning model, including input layer, convolutional layer, fully connected layer and output layer;

[0042] Scan the input data using the convolution kernel;

[0043] The output of the convolutional layer Flatten into a one-dimensional vector c;

[0044] Classify the extracted features through the fully connected layer;

[0045] Substitute the training set into the input layer, get the predicted value through the output of each layer, and complete the forward propagation;

[0046] The loss is calculated using the cross entropy loss function;

[0047] Calculate the loss function for , The gradients of the h, h, and d parameters are used to complete back propagation, and the gradient descent optimizer is used to update the parameters. The steps of forward propagation, loss calculation, back propagation, and parameter update are repeated. The accuracy is calculated through the validation set to calculate the ratio of the number of samples predicted correctly by the model to the total number of samples. When the calculated accuracy reaches the set target accuracy, the model is considered to have reached the predetermined performance standard and the currently updated , , h and d are used as model parameters to complete the model training;

[0048] The complete feature matrix X is input into the model, and different types of pollutants are classified and identified through the deep learning model, and the classification results of various types of pollutants in each monitoring sample are counted;

[0049] The number and proportion of various pollutants in each area are counted, and the pollutant classification results of different monitoring areas are visualized in the form of charts. The geographic information system tool GIS is used to map the pollutant classification results of each monitoring area to the corresponding grid. The degree of pollution is indicated by color coding according to the pollution index. The color gradually changes from blue for low pollution to red for high pollution. The darkest areas in the heat map are identified as pollution hotspots. The pollution hotspots in the monitoring area are displayed in real time based on the input of remote sensing data, and the water quality monitoring of the monitoring area is completed.

[0050] As a preferred solution of the water quality monitoring method based on remote sensing data of the present invention, wherein: the encryption of the classification data and the feature data and uploading them to the cloud server and generating reports includes:

[0051] The complete feature matrix X and classification result data are encrypted using the encryption protocol TLS / SSL and uploaded to the cloud server for storage via a wireless network.

[0052] Based on the uploaded data, staff can remotely download the data through the cloud server and generate water quality monitoring reports through printing devices and using the visualization tool Tableau.

[0053] Another object of the present invention is to provide a system for water quality monitoring method based on remote sensing data, which comprises:

[0054] The data acquisition module is equipped with hyperspectral and multispectral sensors to monitor the area, collect data, and transmit the collected hyperspectral and multispectral data to the ground station;

[0055] The data preprocessing module preprocesses the hyperspectral data and multispectral data, and constructs the corrected hyperspectral data and multispectral data into a unified data set;

[0056] Data registration and fusion module calculates the probability distribution of hyperspectral data and multispectral data, performs spatial registration, performs data fusion based on the probability distribution, and generates a fused data set;

[0057] Feature extraction module, which extracts the main features through principal component analysis and extracts the spectral features in hyperspectral and multispectral data;

[0058] The deep learning classification module trains and tests the model, optimizes the model parameters, uses the trained model to classify new data, and identifies the characteristics of different categories;

[0059] The data transmission module encrypts the classification data and feature data and uploads the encrypted data to the cloud server.

[0060] A computer device comprises: a memory and a processor; the memory stores a computer program, and the processor implements the steps of the water quality monitoring method based on remote sensing data when executing the computer program.

[0061] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the water quality monitoring method based on remote sensing data.

[0062] The beneficial effects of the present invention are as follows: by calculating the probability distribution, the registration of hyperspectral data and multispectral data is achieved, ensuring that they can be compared and analyzed in the same spatial reference system, and by data fusion, the data are made spatially consistent to better reflect the actual condition of the water body. By performing comprehensive feature extraction on the data and building a deep learning model, different categories of pollutants can be classified and identified, and the classification results of various pollutants in each monitoring sample are statistically analyzed to achieve accurate classification. Pollution hotspots can also be marked based on the comparison of the monitoring sample areas, providing reliable water quality assessment results. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0064] Figure 1 Schematic diagram of the process of water quality monitoring method based on remote sensing data.

[0065] Figure 2 Schematic diagram of the water quality monitoring system based on remote sensing data. DETAILED DESCRIPTION

[0066] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings.

[0067] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0068] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or selective embodiment that is mutually exclusive with other embodiments. Example

[0069] Reference Figure 1 , which is the first embodiment of the present invention, and which provides a water quality monitoring method based on remote sensing data. The water quality monitoring method based on remote sensing data includes:

[0070] S1, based on the remote sensing platform, collect hyperspectral data and multispectral data of the monitoring area, form data sets with the collected hyperspectral data and multispectral data, and perform data correction;

[0071] Preferably, collecting hyperspectral data and multispectral data of the monitoring area based on the remote sensing platform includes:

[0072] The remote sensing platform is used to collect data from monitoring areas in different regions. The remote sensing platform is equipped with hyperspectral sensors and multispectral sensors. The hyperspectral sensor is used to obtain hyperspectral data of the monitoring area, ensuring that the spectral resolution of the hyperspectral sensor can cover the 400nm to 2500nm band. The data is stored in the form of cubes. The multispectral sensor is used to obtain multispectral data of the same area, ensuring that the spatial resolution of the multispectral sensor is not less than 10 meters and covers the infrared, near-infrared, and visible light bands. The data is stored in the form of two-dimensional images;

[0073] Collect solar irradiance data, reflectivity data and noise data for the monitoring area, and calculate the distance data between the sun and the monitoring area;

[0074] Download data from the sensor to the ground control station and back up the original data to an external storage device.

[0075] Through the efficient data collection of the remote sensing platform, the time and labor costs of monitoring are reduced. Through the collection of hyperspectral data and multispectral data, data sets are formed respectively, which realizes the comprehensive capture and characterization of water quality information. Hyperspectral sensors can capture detailed spectral information and provide higher resolution spectral data, which helps to finely distinguish and identify different components in the water; multispectral sensors provide the overall spectral characteristics of water quality by capturing information from a few wide spectral bands.

[0076] Furthermore, the collected hyperspectral data and multispectral data are respectively formed into data sets, and data correction is performed including:

[0077] The collected hyperspectral data constitute a hyperspectral dataset , expressed as , where M and N are spatial dimension data, L is the number of hyperspectral bands, for hyperspectral datasets Perform radiation correction, expressed as:

[0078] ,

[0079] in represents the corrected hyperspectral data, represents the solar irradiance, represents the original hyperspectral data, Indicates the distance between the sun and the monitoring area, Represents noise, and through the noise model provided in the technical manual, subtract the noise to obtain the real reflectivity data;

[0080] Use geographic reference points GCPs to match known geographic locations and corresponding pixels in hyperspectral data, perform geometric correction, and convert the corrected hyperspectral data Project it to the standard geographic coordinate system to obtain the preprocessed hyperspectral dataset E;

[0081] The collected multispectral data form a multispectral dataset, which is expressed as , where K is the number of multispectral bands, and the geometric correction is performed using geographic reference points. By matching geographic control points GCPs, the image is corrected to a unified geographic coordinate system, and the multispectral dataset is reprojected to a standard geographic coordinate system to obtain the preprocessed multispectral dataset. ;

[0082] The atmospheric transmittance and atmospheric reflectance data of the monitoring area are obtained, and the preprocessed multispectral data set is atmospherically corrected, which is expressed as:

[0083] ,

[0084] Where T represents the atmospheric transmittance, represents the atmospheric reflectivity, Represented as a geometrically corrected multispectral dataset, Represented as the corrected multispectral dataset.

[0085] Data correction reduces interference from ambient lighting, sensor performance and angle, improves data quality and accuracy, and ensures that the data can truly reflect the spectral characteristics of the water body through steps such as radiation correction, geometric correction and atmospheric correction. Accurate spectral data can better reflect the composition and distribution of substances in the water body, providing a high-quality data source for subsequent data fusion, feature extraction and classification evaluation.

[0086] S2, align the corrected hyperspectral dataset and multispectral dataset by calculating the probability distribution, perform data fusion, and extract fusion features based on principal component features and spectral features;

[0087] Preferably, the probability distribution of the corrected hyperspectral data set and the multispectral data set is calculated and registered, and data fusion is performed, including:

[0088] Create a 2D matrix , the size is , calculate the joint probability distribution, expressed as:

[0089] ,

[0090] in Expressed as a joint probability distribution matrix, Represented as an initialized two-dimensional matrix, each element in the matrix represents the number of data points in a certain interval i and j, i and j represent different intervals of two variables, and They are the gray levels of hyperspectral data and multispectral data, respectively. The data collected by hyperspectral and multispectral imaging systems usually have different gray levels.

[0091] Calculate the marginal probability distribution of hyperspectral data and multispectral data, expressed as:

[0092] ,

[0093] ,

[0094] in and represent the marginal probability distribution of hyperspectral data and the marginal probability distribution of multispectral data respectively;

[0095] Hyperspectral data and multispectral data are registered by maximizing mutual information, which can be expressed as:

[0096] ,

[0097] in is the mutual information, which indicates the degree of information sharing between hyperspectral data and multispectral data;

[0098] Define the spatial transformation parameters including translation, rotation and scaling, and set the parameters to ,in and represents the amount of translation, represents the rotation angle, and is the scaling factor;

[0099] Initialization setting parameters , which means no translation, no rotation and no scaling, and is applied to the hyperspectral data to get the initialized hyperspectral data ,The translation transformation is achieved by moving each pixel position in the data set by a certain distance, the rotation transformation is achieved by rotating each pixel position in the image, and the scaling transformation is achieved by adjusting each pixel position in the image. For the initial parameters, there is no translation, no rotation, and no scaling. Applying these initial transformation parameters to the hyperspectral data E will not make any actual transformation to the data;

[0100] Set the optimization goal to maximize mutual information, expressed as:

[0101] ,

[0102] The gradient descent method is used to perform iterative optimization calculations to gradually increase the mutual information. The minimum value is set as the target threshold based on historical data. The iteration is stopped when the iterative change is less than the target threshold, and the optimal transformation parameters are set. Applied to the hyperspectral dataset E, the spatial alignment of the hyperspectral dataset E and the multispectral dataset I is completed;

[0103] The hyperspectral data set E and the multispectral data set I are respectively subjected to wavelet transform, and processed by discrete wavelet transform DWT. Each band of the hyperspectral data and the multispectral data is subjected to two-dimensional discrete wavelet transform to obtain the corresponding wavelet coefficients. and , and perform data fusion, expressed as:

[0104] ,

[0105] in is the fused data, represents the wavelet coefficients of the k-th layer of hyperspectral data, represents the wavelet coefficient of the k-th layer of multispectral data, and v represents the number of layers of wavelet decomposition;

[0106] Represent the fused data as a three-dimensional matrix , the size is .

[0107] By calculating the probability distribution, the registration of hyperspectral data and multispectral data is achieved, ensuring that different types of data can be compared and analyzed in the same spatial reference system. The fused data is not only highly consistent in space, but also more comprehensive in spectral information, which can better reflect the actual condition of the water body and provide reliable data support for water quality analysis.

[0108] Furthermore, fusion feature extraction based on principal component features and spectral features includes:

[0109] Create an initialized two-dimensional matrix X and convert the three-dimensional matrix Expand and fill the two-dimensional matrix X, where each row represents the spectral information of a pixel, each column represents a band, and the matrix size is expressed as ;

[0110] Remove the mean of the data of each band of the two-dimensional matrix X to form a data matrix after removing the mean , and calculate the mean-removed data matrix The covariance matrix C of , the eigenvalue decomposition of the covariance matrix C is expressed as:

[0111] ,

[0112] ,

[0113] in represents the eigenvalue diagonal matrix, is represented as a matrix of eigenvectors, express The transpose of represents the transpose of W;

[0114] Set the contribution threshold based on historical data, calculate the cumulative variance contribution rate of the selected eigenvalue number d, and determine the number of d when the contribution rate reaches the contribution threshold, and form the eigenvector matrix with the eigenvector corresponding to the selected eigenvalue , the data matrix after removing the mean Projection to the eigenvector matrix , used to remove the mean data matrix With the eigenvector matrix Multiply them together to get the principal component feature matrix Z after dimensionality reduction. The matrix size is expressed as ;

[0115] Each band is normalized from the fused three-dimensional matrix F and the reflectance is calculated, which is expressed as:

[0116] ,

[0117] in Represents each band The normalized reflectivity, represents the radiance of the band in a three-dimensional matrix, and Represents the maximum and minimum radiance, respectively, based on Composition of normalized reflectance matrix ;

[0118] Based on the normalized reflectance matrix T, the point with the lowest reflectance in the spectrum curve is determined as the absorption peak position;

[0119] Based on the normalized reflectance matrix T, for each pair of bands and , calculate the spectral slope, expressed as:

[0120] ,

[0121] in and are the normalized reflectances of the two bands, represents the spectral slope of the band;

[0122] based on Composing the spectral slope matrix , where q is the number of band logarithms;

[0123] The spectral feature matrix is ​​formed based on the normalized reflectance matrix T, the absorption peak position and the spectral slope matrix S , the matrix size is By combining the normalized reflectance matrix R, the absorption peak position and the spectral slope matrix S, the characteristics of the spectral data can be fully described. The normalized reflectance provides basic spectral reflectance information, the absorption peak position provides the key point for material identification, and the spectral slope provides detailed shape characteristics. Through the combination of multiple features, the distinguishing ability of the spectral feature matrix can be significantly improved.

[0124] PCA can capture the main change patterns in the data and achieve the effect of extracting the main component features. Spectral feature extraction is to extract spectral features that can represent the water quality status by analyzing the spectral information of the data. These features can reflect the various components in the water and provide a basis for water quality assessment. Through feature extraction, the dimension and complexity of the data are reduced. The combination of principal component features and spectral features can not only capture the main change patterns in the data, but also highlight the spectral features that reflect the water quality, thereby improving the accuracy and reliability of data analysis, making the processed data more refined and accurate, providing a reliable basis for water quality assessment, and improving the performance and effectiveness of the entire monitoring system.

[0125] S3, build a deep learning model to classify data based on data fusion features and evaluate the water quality in the monitoring area;

[0126] Preferably, a deep learning model is constructed to classify data based on data fusion features, and the water quality of the monitored area is evaluated including:

[0127] The principal component feature matrix Z and the spectral feature matrix after dimensionality reduction are Combine into a complete feature matrix X;

[0128] Build a deep learning model, including input layer, convolutional layer, fully connected layer and output layer;

[0129] Use the convolution kernel to scan the input data, expressed as:

[0130] ,

[0131] in Represents the output of the convolutional layer, ReLU is the activation function, K represents the total number of convolution kernels, represents the kth convolution kernel, * represents the convolution operation, Represents the bias of the kth convolution kernel;

[0132] The output of the convolutional layer Flatten into a one-dimensional vector c;

[0133] The extracted features are classified through the fully connected layer, expressed as:

[0134] ,

[0135] Where h is the weight matrix, d is the bias term, and soft is the softmax activation function. Represents the output of the fully connected layer;

[0136] Substitute the training set into the input layer, get the predicted value through the output of each layer, and complete the forward propagation;

[0137] The loss is calculated using the cross entropy loss function, expressed as:

[0138] ,

[0139] Where N is the number of samples, G is the number of categories, represents the true label of the i-th g class, represents the predicted probability of the ith c class, Indicates calculated loss;

[0140] Calculate the loss function for , The gradients of the h, h, and d parameters are used to complete back propagation, and the gradient descent optimizer is used to update the parameters. The steps of forward propagation, loss calculation, back propagation, and parameter update are repeated. The accuracy is calculated through the validation set to calculate the ratio of the number of samples predicted correctly by the model to the total number of samples. When the calculated accuracy reaches the set target accuracy, the model is considered to have reached the predetermined performance standard and the currently updated , , h and d are used as model parameters to complete the model training;

[0141] The complete feature matrix X is input into the model, and different types of pollutants are classified and identified through the deep learning model, and the classification results of various types of pollutants in each monitoring sample are counted;

[0142] The number and proportion of various pollutants in each area are counted, and the pollutant classification results of different monitoring areas are visualized in the form of charts. The geographic information system tool GIS is used to map the pollutant classification results of each monitoring area to the corresponding grid. Color coding is used to indicate the degree of pollution according to the pollution index. The color gradually changes from blue for low pollution to red for high pollution. The darkest areas in the heat map are identified as pollution hotspots. The pollution hotspots in the monitoring area are displayed in real time based on the input of remote sensing data. The water quality monitoring of the monitoring area is completed. Combined with the location of the hotspots in the heat map, correlation analysis can be performed with geographic information, land use, industrial layout and other information to find out potential sources of pollution.

[0143] By training the model on a large amount of historical data, the model can automatically learn the complex patterns and changing trends in the data, thereby achieving more accurate water quality classification. Through the classification results of the deep learning model, the classification results of various pollutants in each monitoring sample can be statistically analyzed. By comparing the pollutant classification results in different monitoring areas, pollution hotspots can be marked, and water quality problems can be discovered and dealt with in a timely manner. This process improves the efficiency and accuracy of water quality monitoring.

[0144] S4, encrypts the classification data and feature data and uploads them to the cloud server and generates reports;

[0145] Preferably, encrypting the classification data and feature data and uploading them to the cloud server and generating reports includes:

[0146] The complete feature matrix X and classification result data are encrypted using the encryption protocol TLS / SSL and uploaded to the cloud server for storage via the wireless network.

[0147] Based on the uploaded data, staff can remotely download the data through the cloud server and generate water quality monitoring reports through printing devices and using the visualization tool Tableau.

[0148] Data encryption and cloud storage ensure data security and convenience, protect data privacy and security, and prevent unauthorized access and use. The application of cloud servers makes data storage and management more efficient and convenient, ensuring data security and reliability.

[0149] Example 2

[0150] Reference Figure 2 , which is the second embodiment of the present invention, and which is different from the previous embodiment, provides a system for a water quality monitoring method based on remote sensing data, including:

[0151] The data acquisition module is equipped with hyperspectral and multispectral sensors to monitor the area, collect data, and transmit the collected hyperspectral and multispectral data to the ground station;

[0152] The data preprocessing module preprocesses the hyperspectral data and multispectral data, and constructs the corrected hyperspectral data and multispectral data into a unified data set;

[0153] Data registration and fusion module calculates the probability distribution of hyperspectral data and multispectral data, performs spatial registration, performs data fusion based on the probability distribution, and generates a fused data set;

[0154] Feature extraction module, which extracts the main features through principal component analysis and extracts the spectral features in hyperspectral and multispectral data;

[0155] The deep learning classification module trains and tests the model, optimizes the model parameters, uses the trained model to classify new data, and identifies the characteristics of different categories;

[0156] The data transmission module encrypts the classification data and feature data and uploads the encrypted data to the cloud server.

[0157] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0158] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0159] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0160] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit with a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit with a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0161] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A water quality monitoring method based on remote sensing data, Features: include, Based on the remote sensing platform, the hyperspectral data and multispectral data of the monitoring area are collected, and the collected hyperspectral data and multispectral data are respectively formed into data sets, and data correction is performed; The probability distribution of the corrected hyperspectral dataset and multispectral dataset is calculated and registered, and data fusion is performed, and fusion features are extracted based on principal component features and spectral features; Build a deep learning model to classify data based on data fusion features and evaluate the water quality in the monitoring area; Encrypt the classification data and feature data and upload them to the cloud server to generate reports; The hyperspectral data and multispectral data of the monitoring area collected based on the remote sensing platform include: The remote sensing platform is used to collect data from monitoring areas in different regions. The remote sensing platform is equipped with hyperspectral sensors and multispectral sensors. The hyperspectral sensors are used to obtain hyperspectral data of the monitoring areas, and the multispectral sensors are used to obtain multispectral data of the same area. Collect solar irradiance data, reflectivity data and noise data for the monitoring area, and calculate the distance data between the sun and the monitoring area; Download data from sensors to the ground control station and back up raw data to external storage devices; The collected hyperspectral data and multispectral data are respectively formed into data sets, and data correction is performed, including: The collected hyperspectral data constitute a hyperspectral dataset ; Use geographic reference points to perform geometric correction and convert the corrected hyperspectral data Project it to the standard geographic coordinate system to obtain the preprocessed hyperspectral dataset E; The collected multispectral data form a multispectral dataset, which is expressed as , where M and N are spatial dimension data, K is the number of multispectral bands, and the multispectral dataset is reprojected to the standard geographic coordinate system using geographic reference points for geometric correction to obtain the preprocessed multispectral dataset ; Obtain atmospheric transmittance and atmospheric reflectance data of the monitoring area, and perform atmospheric correction on the preprocessed multispectral data set; The calculation of probability distribution of the corrected hyperspectral data set and multispectral data set is performed to register, and data fusion is performed, including: Create a two-dimensional matrix; Calculate marginal probability distributions for hyperspectral and multispectral data; Hyperspectral data and multispectral data are registered by maximizing mutual information; Define the spatial transformation parameters including translation, rotation and scaling, and set the parameters to ,in and represents the amount of translation, represents the rotation angle, and is the scaling factor; Initialization setting parameters , which means no translation, no rotation and no scaling, and is applied to the hyperspectral data to get the initialized hyperspectral data ; Set the optimization goal to maximize mutual information, The gradient descent method is used to perform iterative optimization calculations, and the minimum value is set as the target threshold based on historical data. The iteration is stopped when the iterative change is less than the target threshold, and the optimal transformation parameters are set. Applied to the hyperspectral dataset E, the spatial alignment of the hyperspectral dataset E and the multispectral dataset I is completed; Perform wavelet transform on the hyperspectral dataset E and multispectral dataset I respectively to obtain the wavelet coefficients and , and perform data fusion; Represent the fused data as a three-dimensional matrix F; The fusion feature extraction based on the principal component feature and the spectral feature includes: Create an initialized two-dimensional matrix X and convert the three-dimensional matrix Expand and fill the two-dimensional matrix X, the matrix size is expressed as ; Remove the mean of the data of each band of the two-dimensional matrix X to form a data matrix after removing the mean , and calculate the mean-removed data matrix The covariance matrix C of , performs eigenvalue decomposition on the covariance matrix C; Set the contribution threshold based on historical data, calculate the cumulative variance contribution rate of the selected eigenvalue number d, and determine the number of d when the contribution rate reaches the contribution threshold, and form the eigenvector matrix with the eigenvector corresponding to the selected eigenvalue , the data matrix after removing the mean Projection to the eigenvector matrix , we get the principal component feature matrix Z after dimensionality reduction, and the matrix size is expressed as ; Normalize each band from the fused three-dimensional matrix F and calculate the reflectance; Based on the normalized reflectance matrix T, the point with the lowest reflectance in the spectrum curve is determined as the absorption peak position; Based on the normalized reflectance matrix T, for each pair of bands and , calculate the spectral slope; based on Composing the spectral slope matrix , where q is the number of band logarithms; The spectral feature matrix is ​​formed based on the normalized reflectance matrix T, the absorption peak position and the spectral slope matrix S , the matrix size is .

2. The water quality monitoring method based on remote sensing data as claimed in claim 1, Features: The deep learning model is constructed to classify data based on data fusion features, and the water quality of the monitoring area is evaluated, including: The principal component feature matrix Z and the spectral feature matrix after dimensionality reduction are Combine into a complete feature matrix X; Build a deep learning model, including input layer, convolutional layer, fully connected layer and output layer; Scan the input data using the convolution kernel; The output of the convolutional layer Flatten into a one-dimensional vector c; Classify the extracted features through the fully connected layer; Substitute the training set into the input layer, get the predicted value through the output of each layer, and complete the forward propagation; The loss is calculated using the cross entropy loss function; Calculate the loss function for , The gradients of the h, h, and d parameters are used to complete back propagation, and the gradient descent optimizer is used to update the parameters. The steps of forward propagation, loss calculation, back propagation, and parameter update are repeated. The accuracy is calculated through the validation set to calculate the ratio of the number of samples predicted correctly by the model to the total number of samples. When the calculated accuracy reaches the set target accuracy, the model is considered to have reached the predetermined performance standard and the currently updated , , h and d are used as model parameters to complete the model training; The complete feature matrix X is input into the model, and different types of pollutants are classified and identified through the deep learning model, and the classification results of various types of pollutants in each monitoring sample are counted; The number and proportion of various pollutants in each area are counted, and the pollutant classification results of different monitoring areas are visualized in the form of charts. The geographic information system tool GIS is used to map the pollutant classification results of each monitoring area to the corresponding grid. The degree of pollution is indicated by color coding according to the pollution index. The color gradually changes from blue for low pollution to red for high pollution. The darkest areas in the heat map are identified as pollution hotspots. The pollution hotspots in the monitoring area are displayed in real time based on the input of remote sensing data, and the water quality monitoring of the monitoring area is completed.

3. The water quality monitoring method based on remote sensing data as claimed in claim 2, Features: The encrypted uploading of the classification data and feature data to the cloud server and generating reports includes: The complete feature matrix X and classification result data are encrypted using the encryption protocol TLS / SSL and uploaded to the cloud server for storage via the wireless network. Based on the uploaded data, staff can remotely download the data through the cloud server and generate water quality monitoring reports through printing devices and using the visualization tool Tableau.

4. A system based on the water quality monitoring method based on remote sensing data according to any one of claims 1 to 3, Features: include, The data acquisition module is equipped with hyperspectral and multispectral sensors to monitor the area, collect data, and transmit the collected hyperspectral and multispectral data to the ground station; The data preprocessing module preprocesses the hyperspectral data and multispectral data, and constructs the corrected hyperspectral data and multispectral data into a unified data set; Data registration and fusion module calculates the probability distribution of hyperspectral data and multispectral data, performs spatial registration, performs data fusion based on the probability distribution, and generates a fused data set; Feature extraction module, which extracts the main features through principal component analysis and extracts the spectral features in hyperspectral and multispectral data; The deep learning classification module trains and tests the model, optimizes the model parameters, uses the trained model to classify new data, and identifies the features of different categories; The data transmission module encrypts the classification data and feature data and uploads the encrypted data to the cloud server.

5. A computer device, include: Memory and processor; The memory stores a computer program, characterized in that: when the processor executes the computer program, the steps of the water quality monitoring method based on remote sensing data described in any one of claims 1 to 3 are implemented.

6. A computer-readable storage medium having a computer program stored thereon, Features: When the computer program is executed by a processor, the steps of the water quality monitoring method based on remote sensing data described in any one of claims 1 to 3 are implemented.

Images