River water quality monitoring method and system based on hyperspectral remote sensing of unmanned aerial vehicle

Abstract

The invention relates to the technical field of optical line protection, and discloses a river water quality monitoring method based on an unmanned aerial vehicle hyperspectral remote sensing image and machine learning, which combines the unmanned aerial vehicle hyperspectral remote sensing image with a machine learning algorithm to construct a river water quality parameter inversion model. Comprising the following steps: collecting a water sample, obtaining an unmanned aerial vehicle hyperspectral image and preprocessing; extracting an image of the river part and selecting the reflectivity of each wave band of the image; performing spectrum transformation on the original spectrum of the image, establishing an inversion data set through feature screening, and constructing a water quality parameter inversion model by using a machine learning algorithm; and simulating the water quality parameters of the river through the water quality parameter inversion model, and visualizing the simulation result. According to the method, urban inland small river water quality monitoring can be rapidly achieved at low cost, the dual problem that the time resolution and the space resolution are limited in urban small river monitoring based on satellite remote sensing images is solved, and a scientific basis is provided for urban river intelligent inversion.

Description

Technical Field

[0001] The present invention relates to the technical field of optical line protection, and in particular to a river water quality monitoring method and system based on unmanned aerial vehicle hyperspectral remote sensing. Background Art

[0002] At present, there are two main methods for monitoring the water quality of surface water bodies: ground detection and remote sensing monitoring. Ground monitoring includes manual field sampling and automatic site monitoring. This type of method can effectively detect the information of collected water samples and has the advantages of high monitoring accuracy and strong research pertinence, but the process is complicated, time-consuming and labor-intensive. And it cannot meet the requirements of real-time dynamic watershed-scale water quality monitoring. In recent years, remote sensing monitoring technology has begun to be applied to water body research. Remote sensing satellites can achieve large-area synchronous observations and can be used to monitor and analyze the spatial distribution characteristics of water quality in the entire river. However, due to the limitations of time and spatial resolution, it is difficult to meet the data requirements of high spatial resolution and short return period required for inland river water quality monitoring.

[0003] As technology continues to mature, drones can take aerial photos from all directions and angles with low manpower consumption, which provides a new idea for the dynamic assessment of water quality in inland small watersheds. At the same time, with the continuous advancement of remote sensing technology, hyperspectral sensors can obtain continuous spectral data through each pixel, opening up a new way for dynamic monitoring of water quality in the entire region. Drones equipped with hyperspectral sensors can not only give full play to the advantages of drones' maneuverability and flexibility, but also take into account the performance of hyperspectral equipment, conduct an overall scan of the target object, establish a relationship model between spectral characteristics and water quality indicators, and effectively invert water quality indicators. Summary of the invention

[0004] The purpose of the present invention is to provide a river water quality monitoring method and system based on unmanned aerial vehicle hyperspectral remote sensing to solve the above technical problems.

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] A method for monitoring river water quality based on unmanned aerial vehicle hyperspectral remote sensing, the method comprising the following steps:

[0007] S1: Set up sampling points in the study area to collect water samples, use GPS positioning measuring instruments to obtain the latitude and longitude of the sampling points, and after completing the water sampling of all sampling points, send them to the laboratory for chemical analysis of water quality parameters. While collecting water samples, use a drone equipped with a hyperspectral imager to obtain hyperspectral images of the corresponding study area;

[0008] S2: Extract the band reflectance of the location in the image according to the longitude and latitude of the artificial sampling point;

[0009] S3: performing different forms of spectral transformation on the original spectral reflectance in step S2 to form a water quality inversion data set;

[0010] S4: Based on the water quality inversion data set in step S3 above, a water quality parameter inversion model is constructed using a machine learning algorithm;

[0011] S5: The water quality parameters of the river are simulated through the water quality parameter inversion model, and the simulation results are visualized to show the spatial distribution characteristics of the water quality parameter concentrations.

[0012] As a further description of the scheme of the present invention, in S1, acquiring hyperspectral images means: using a drone equipped with an airborne hyperspectral imager to formulate a detailed flight plan before the flight, wherein the flight plan includes route planning, flight altitude, and camera triggering interval.

[0013] As a further description of the scheme of the present invention, in S1, hyperspectral image preprocessing refers to: using specview software and arcgis software to preprocess the drone hyperspectral image, and the preprocessing process includes four steps: lens calibration, reflectivity calibration, atmospheric correction, and image stitching.

[0014] As a further description of the scheme of the present invention, the specific method of S2 includes: in the ENV I software, constructing an ROI region of interest centered on the sampling point, and taking the average spectral reflectance of all points in the ROI region of interest as the spectral data of the water quality sampling points in the area.

[0015] As a further description of the scheme of the present invention, in S3, MATLAB software is used to perform spectrum transformation on the original spectrum, and the transformation method includes SG smoothing, normalization processing and fractional differential processing.

[0016] As a further description of the scheme of the present invention, in S3, the formation of the water quality inversion data set refers to using the Pearson correlation coefficient method to extract features from the water quality hyperspectral data after spectral transformation, and screening out sensitive bands with high correlation to form a data set.

[0017] As a further description of the solution of the present invention, the specific process of S4 includes:

[0018] The BP neural network is used to establish a water quality inversion model. By initializing the network parameters, the single band or band combination data with high correlation with water quality parameters is used as input, and the water quality parameter concentration at the sampling point is used as output. The network is constructed, the output value of the network is calculated, and it is compared with the expected value. If there is an error, the error back propagation algorithm is used to correct the error layer by layer from the output layer to the input layer. By continuously adjusting the connection weights between layers and the thresholds of nodes, the output of the network is made closer to the expected output, and the determination coefficient R is used to2 , root mean square error RMSE and mean absolute percentage error MAE to evaluate the model stability.

[0019] As a further description of the scheme of the present invention, the specific process in S5 includes: obtaining the inversion results of each water quality parameter based on the constructed inversion model, and displaying the spatial distribution characteristics of the concentration of each water quality parameter through layered coloring.

[0020] A river water quality monitoring system based on unmanned aerial vehicle hyperspectral remote sensing, the system is used to implement the river water quality monitoring method based on unmanned aerial vehicle hyperspectral remote sensing.

[0021] Beneficial effects of the present invention:

[0022] The present invention obtains the hyperspectral image of river water body by aerial photography with a hyperspectrometer mounted on an unmanned aerial vehicle, and processes it into a water body reflectance image through an image preprocessing system; collects river water samples and performs chemical analysis, performs multiple spectral transformations on the original hyperspectral data, reduces noise, enhances spectral response, highlights spectrum characteristics, performs correlation analysis on water quality parameters and spectral band reflectance, and selects sensitive bands with higher correlation coefficients. The characteristic bands of each water quality parameter obtained by screening under different transformed spectra are used as model input variables to establish a water quality inversion model, and the water body reflectance image of the unmanned aerial vehicle is imported into the trained water quality inversion model. Finally, a water quality index concentration distribution map is obtained to realize rapid monitoring of water quality in small watersheds. In areas with complex water body components and narrow waters such as urban rivers, the remote sensing data of unmanned aerial vehicles combined with the measured water quality data of a small number of sample points can obtain rich and comprehensive water quality conditions and spatial distribution, providing a scientific basis for the intelligent inversion of urban rivers. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The present invention will be further described below in conjunction with the accompanying drawings.

[0024] Figure 1 It is a partial flow chart of the river water quality monitoring method based on UAV hyperspectral remote sensing provided by the present invention. DETAILED DESCRIPTION

[0025] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] See also Figure 1 As shown, the present invention is a river water quality monitoring method based on UAV hyperspectral remote sensing, and the method comprises the following steps:

[0027] S1: Set up sampling points in the study area to collect water samples, use GPS positioning measuring instruments to obtain the latitude and longitude of the sampling points, and after completing the water sampling of all sampling points, send them to the laboratory for chemical analysis of water quality parameters. While collecting water samples, use a drone equipped with a hyperspectral imager to obtain hyperspectral images of the corresponding study area;

[0028] Acquiring hyperspectral images means: using a drone equipped with an airborne hyperspectral imager to formulate a detailed flight plan before the flight, wherein the flight plan includes route planning, flight altitude, and camera triggering interval.

[0029] Hyperspectral image preprocessing refers to: using specview software and arcgis software to preprocess the UAV hyperspectral images. The preprocessing process includes four steps: lens calibration, reflectivity calibration, atmospheric correction, and image stitching.

[0030] In the above step S1, sampling points are arranged in the study area for water sampling (one point is collected every 50-100m), and the latitude and longitude of the sampling points are obtained using a GPS positioning measuring instrument. After the water samples of all sampling points are collected, they are sent to the laboratory as soon as possible to complete the chemical analysis of water quality parameters. While collecting water samples, a hyperspectral imager is used to obtain hyperspectral images of the corresponding study area using a drone;

[0031] Then, the UAV hyperspectral images were preprocessed using Specview software and ArcGIS software, including four steps: lens calibration, reflectance calibration, atmospheric correction, and image stitching, to obtain UAV hyperspectral images with real surface reflectance.

[0032] The specific implementation method is to select Analysis Tools-Lens-Import Data-Calculate in the toolbar in Specview software to obtain the data that has been calibrated for reflectivity; select Analysis Tools-Reflectivity Calibration-Import Data-Calculate in the toolbar to obtain the accurate reflectivity number; complete the atmospheric correction in the toolbar-Atmosphere Correction. Import the standard data remote sensing image into ArcGIS and stitch several images into a complete remote sensing image.

[0033] Finally, the location of the water body in the study area was determined based on field surveys and satellite images of the study area. The water body boundary of the UAV hyperspectral image was outlined using ArcGIS to obtain the vector boundary of the river. The mask extraction function in ArcGIS software was used to extract the UAV hyperspectral image corresponding to the study river.

[0034] It should be noted that airborne hyperspectral data acquisition uses a drone equipped with a hyperspectral imager to acquire image data in the study area, and continuously images the target area in hundreds of spectral bands within the spectral coverage range. The time selected for the drone flight is clear and breezy (level 1-2) with no cloud cover.

[0035] S2: Extract the band reflectance of the location in the image according to the longitude and latitude of the artificial sampling point;

[0036] The specific method of S2 includes: in the ENV I software, constructing an ROI region of interest centered on the sampling point, and taking the average spectral reflectance of all points in the ROI region of interest as the spectral data of the water quality sampling points in the area.

[0037] S3: performing different forms of spectral transformation on the original spectral reflectance in step S2 to form a water quality inversion data set;

[0038] In S3, MATLAB software is used to perform spectrum transformation on the original spectrum, and the transformation method includes SG smoothing, normalization processing and fractional differential processing.

[0039] It should be noted that SG smoothing uses a polynomial fitting method to smooth the spectral data. It performs a sliding window operation on the data, performs a polynomial fitting on the data points in each window, and then replaces the original data points with the values ​​on the fitting curve to achieve a smoothing effect. This method can effectively remove noise and stray signals in the spectrum and retain the main features of the spectrum.

[0040] Spectral normalization is to scale and translate the spectral data proportionally so that the data falls into a small specific interval, eliminating the impact of the data dimension and making the data indicators comparable.

[0041] The fractional order differential of the spectrum is a generalization of the traditional integer order differential. It can handle non-integer order differentials, thus providing more flexible feature extraction and data processing capabilities. The use of fractional order differential spectra can highlight the hidden information in the band, improve the sensitivity of spectral information, and enhance the ability to analyze data.

[0042] In S3, the formation of the water quality inversion data set refers to extracting features from the water quality hyperspectral data after spectral transformation using the Pearson correlation coefficient method, and screening out sensitive bands with high correlation to form a data set.

[0043] SPSS software was used to extract features from the water quality hyperspectral data after spectral transformation using the Pearson correlation coefficient (PCC) method. The Pearson correlation coefficient between each wavelength and the water quality index was calculated. The Pearson correlation coefficient is used to measure the linear correlation between two variables, and its value range is [-1,1], where 1 indicates a complete positive correlation, -1 indicates a complete negative correlation, and 0 indicates no correlation. Generally speaking, a wavelength with an absolute value of a correlation coefficient greater than 0.6 can be considered to be significantly correlated with the water quality index.

[0044] It should be noted that the person correlation coefficient method is used to perform correlation analysis on the original spectral data after spectral transformation with the water quality parameter data of the corresponding points, and the sensitive bands with higher correlation coefficients are selected.

[0045] In SPSS, after opening the data file, select Analysis-Correlation-Bivariate in the toolbar to output the results, view the value of the correlation coefficient and its significance level, and select the characteristic band with higher correlation as the predictive variable of water quality parameters based on the size and significance level of the correlation coefficient.

[0046] S4: Based on the water quality inversion data set in step S3 above, a water quality parameter inversion model is constructed using a machine learning algorithm;

[0047] The specific process of S4 includes:

[0048] The BP neural network is used to establish a water quality inversion model. By initializing the network parameters, the single band or band combination data with high correlation with water quality parameters is used as input, and the water quality parameter concentration at the sampling point is used as output. The network is constructed, the output value of the network is calculated, and it is compared with the expected value. If there is an error, the error back propagation algorithm is used to correct the error layer by layer from the output layer to the input layer. By continuously adjusting the connection weights between layers and the thresholds of nodes, the output of the network is made closer to the expected output, and the determination coefficient R is used to 2 , root mean square error RMSE and mean absolute percentage error MAE to evaluate the model stability.

[0049] The specific steps of the BP neural network algorithm are as follows:

[0050] Step 1: Initialize the connection threshold and weight, and set the value to a relatively small random number;

[0051] Step 2: Input training samples;

[0052] Step 3: Calculate the actual output of each node in the input layer and hidden layer, and calculate the local error of all layers;

[0053] Step 4: Correct the thresholds and weights of the output layer and hidden layer.

[0054] Step 5: Determine whether the output value meets expectations. If it meets expectations, it will be directly output. If it does not meet expectations, it will be transmitted in the reverse direction and reach the hidden layer again. Each node in the hidden layer receives the error signal and adjusts the weights and thresholds of the neural nodes in each layer. Through back propagation, the data is repeatedly trained to obtain a neural network.

[0055] S5: The water quality parameters of the river are simulated through the water quality parameter inversion model, and the simulation results are visualized to show the spatial distribution characteristics of the water quality parameter concentrations.

[0056] The specific process in S5 includes: obtaining the inversion results of various water quality parameters based on the constructed inversion model, and displaying the spatial distribution characteristics of the concentration of each water quality parameter through layered coloring.

[0057] The UAV hyperspectral water quality parameter inversion model built based on BP neural network can invert the inversion results of various water quality parameters, that is, each grid unit in the hyperspectral image can correspond to a water quality parameter concentration value. In arcg is, the inversion results are layered and colored, and appropriate color bands are selected to express different levels of water quality parameter concentration. The spatial distribution characteristics of water quality parameter concentration can be intuitively displayed.

[0058] A river water quality monitoring system based on unmanned aerial vehicle hyperspectral remote sensing, the system is used to implement the river water quality monitoring method based on unmanned aerial vehicle hyperspectral remote sensing.

[0059] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A river water quality monitoring method based on unmanned aerial vehicle hyperspectral remote sensing, characterized in that: The method comprises the following steps: S1: Set up sampling points in the study area to collect water samples, use GPS positioning measuring instruments to obtain the latitude and longitude of the sampling points, and after completing the water sampling of all sampling points, send them to the laboratory for chemical analysis of water quality parameters. While collecting water samples, use a drone equipped with a hyperspectral imager to obtain hyperspectral images of the corresponding study area; S2: Extract the band reflectance of the location in the image according to the longitude and latitude of the artificial sampling point; S3: performing different forms of spectral transformation on the original spectral reflectance in step S2 to form a water quality inversion data set; S4: Based on the water quality inversion data set in step S3 above, a water quality parameter inversion model is constructed using a machine learning algorithm; S5: The water quality parameters of the river are simulated through the water quality parameter inversion model, and the simulation results are visualized to show the spatial distribution characteristics of the water quality parameter concentrations.

2. The method for river water quality monitoring based on UAV hyperspectral remote sensing according to claim 1 is characterized in that: In S1, acquiring hyperspectral images refers to: using a drone equipped with an airborne hyperspectral imager to formulate a detailed flight plan before the flight, wherein the flight plan includes route planning, flight altitude, and camera triggering interval.

3. The method for river water quality monitoring based on UAV hyperspectral remote sensing according to claim 1 is characterized in that: In S1, hyperspectral image preprocessing refers to: using specview software and arcgis software to preprocess the drone hyperspectral image, and the preprocessing process includes four steps: lens calibration, reflectivity calibration, atmospheric correction, and image stitching.

4. The method for monitoring river water quality based on UAV hyperspectral remote sensing according to claim 3 is characterized in that: The specific method of S2 includes: in the ENV I software, constructing a ROI region of interest centered on the sampling point, and taking the average spectral reflectance of all points in the ROI region of interest as the spectral data of the water quality sampling point in the area.

5. The method for river water quality monitoring based on UAV hyperspectral remote sensing according to claim 1 is characterized in that: In S3, MATLAB software is used to perform spectrum transformation on the original spectrum, and the transformation method includes SG smoothing, normalization processing and fractional differential processing.

6. The method for river water quality monitoring based on UAV hyperspectral remote sensing according to claim 5 is characterized in that: In S3, the formation of the water quality inversion data set refers to extracting features from the water quality hyperspectral data after spectral transformation using the Pearson correlation coefficient method, and screening out sensitive bands with high correlation to form a data set.

7. The method for river water quality monitoring based on UAV hyperspectral remote sensing according to claim 1 is characterized in that: The specific process of S4 includes: The BP neural network is used to establish a water quality inversion model. By initializing the network parameters, the single band or band combination data with high correlation with water quality parameters is used as input, and the water quality parameter concentration at the sampling point is used as output. The network is constructed, the output value of the network is calculated, and it is compared with the expected value. If there is an error, the error back propagation algorithm is used to correct the error layer by layer from the output layer to the input layer. By continuously adjusting the connection weights between layers and the thresholds of nodes, the output of the network is made closer to the expected output, and the determination coefficient R is used to 2 , root mean square error RMSE and mean absolute percentage error MAE to evaluate the model stability.

8. The method for river water quality monitoring based on UAV hyperspectral remote sensing according to claim 1 is characterized in that: The specific process in S5 includes: obtaining the inversion results of various water quality parameters based on the constructed inversion model, and displaying the spatial distribution characteristics of the concentration of each water quality parameter through layered coloring.

9. A river water quality monitoring system based on UAV hyperspectral remote sensing, characterized in that: The system is used to implement the river water quality monitoring method based on unmanned aerial vehicle hyperspectral remote sensing as described in any one of claims 1-8.

Images