A water quality monitoring method and computer program product based on hyperspectral multi-parameters

Abstract

The present invention belongs to a water quality monitoring method. In order to solve the technical problems that the traditional water quality monitoring method must contact the water body, has poor real-time performance, and has high requirements for personnel, a water quality monitoring method based on hyperspectral multi-parameters and a computer program product are provided. The correlation coefficient between the spectral curve and the standard water body spectral curve is calculated, and part of the spectral curve in the hyperspectral image is eliminated according to the correlation coefficient, which can effectively improve the spectral data abnormality caused by environmental factors such as floating objects or bubbles, thereby improving the accuracy of water quality monitoring. Then, the spectral curves that have been corrected and the corresponding water quality parameter concentrations under different water quality parameter concentrations are modeled to obtain a modeling model. When the water quality is subsequently monitored, the corrected spectral curve only needs to be input into the modeling model to accurately and efficiently obtain the water quality parameter concentration, so that the method of the present invention is accurate and efficient.

Description

Technical Field

[0001] The present invention belongs to a water quality monitoring method, and in particular relates to a water quality monitoring method based on hyperspectral multi-parameters and a computer program product. Background Art

[0002] In recent years, people have paid more and more attention to environmental issues. Since the water area on the earth has reached a coverage rate of more than 70%, while fresh water resources only account for 2.7% of it, coupled with the increasingly serious problem of water pollution, the protection of water resources is particularly important.

[0003] Water quality monitoring technology is a key technology for preventing water pollution. With the development of science and technology, water quality monitoring technology has become increasingly mature. At present, the main monitoring methods include handheld portable instrument monitoring method, laboratory chemical method, cabinet measurement method, etc. However, traditional methods still have many disadvantages, such as the need to contact the water body, poor real-time performance, the need for certain professional knowledge, and high requirements for personnel. Summary of the invention

[0004] The present invention provides a water quality monitoring method based on hyperspectral multi-parameters and a computer program product to solve the technical problems of traditional water quality monitoring methods, such as the need to contact water bodies, poor real-time performance and high requirements on personnel.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A water quality monitoring method based on hyperspectral multi-parameters, which is special in that it includes the following steps:

[0007] S1, obtain the modeling dataset

[0008] S1.1, decompose the hyperspectral image reflected by the water surface into a spectral curves along the row direction, where a is the number of pixel rows of the spectrometer detector;

[0009] S1.2, respectively calculate the correlation coefficient R between a spectral curve and the standard water spectral curve, and set a threshold T. If R < T, the corresponding spectral curve is eliminated, otherwise, the corresponding spectral curve is retained to obtain n spectral curves, where n is the number of retained spectral curves, n ≤ a;

[0010] S1.3, calculating the average of n spectral curves, and taking the average as the final spectral curve S of the water body;

[0011] S1.4, performing black and white plate correction on the spectrum curve S to obtain a corrected reflection spectrum curve S';

[0012] S1.5, for water bodies with different water quality parameter concentrations, repeat steps S1.1 to S1.4 to obtain the reflection spectrum curves S' corresponding to the water bodies with m water quality parameter concentrations, forming the modeling data set M:

[0013] M=[S′1,…,S′ m ]

[0014] Where m is an integer greater than 1, S′1 is the reflectance spectrum curve corresponding to the water body with the first water quality parameter concentration, S′ m is the reflectance spectrum curve corresponding to the water body with the mth water quality parameter concentration;

[0015] S2, get the modeling model

[0016] Modeling the reflectance spectrum curves in the modeling data set M and the water quality parameter concentrations corresponding to each reflectance spectrum curve to obtain a modeling model;

[0017] S3, monitor water quality

[0018] The hyperspectral image reflected by the water surface is processed through steps S1.1 to S1.4 and then input into the modeling model to obtain the corresponding water quality parameter concentration.

[0019] Furthermore, in step S1.2, the correlation coefficient R between the a spectral curves and the standard water spectral curve is calculated respectively, specifically:

[0020] A multidimensional spectral curve is extracted from each spectral curve according to the response peak position of each wavelength. The dimension of the multidimensional spectral curve is the number of bands of the hyperspectral image. The correlation coefficient R between the spectral curve a and the standard water spectral curve is calculated by the following formula:

[0021]

[0022] Where X and Y represent the two spectral curves for calculating the correlation coefficient R, Cov(X,Y) is the covariance of X and Y, and Var[X] and Var[Y] are the variances of X and Y, respectively.

[0023] Furthermore, in step S1.5, the water quality parameters include supplemental chlorophyll a, total nitrogen, total phosphorus, transparency, chemical oxygen demand, turbidity, suspended solids and permanganate index.

[0024] Furthermore, in step S1.5, the water bodies with different water quality parameter concentrations are specifically water bodies with different water quality parameter concentrations distributed in a gradient.

[0025] Furthermore, step S2 is specifically to perform modeling through partial least squares PLS or back propagation neural network BPNN.

[0026] Furthermore, the hyperspectral image is specifically acquired by a hyperspectral imager mounted on the smart pole.

[0027] The present invention also provides a computer program product, including a computer program, the special feature of which is that when the program is executed by a processor, the steps of the above-mentioned water quality monitoring method based on hyperspectral multi-parameters are implemented.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] 1. In the water quality monitoring method based on hyperspectral multi-parameters of the present invention, the correlation coefficient between the spectral curve and the standard water body spectral curve is calculated, and part of the spectral curve in the hyperspectral image is eliminated according to the correlation coefficient, which can effectively improve the spectral data abnormality caused by environmental factors such as floating objects or bubbles, thereby improving the accuracy of water quality monitoring. Then, the corrected spectral curve and the corresponding water quality parameter concentration under different water quality parameter concentrations are modeled by PLS or BPNN and other methods to obtain a modeling model. When the water quality is subsequently monitored, the corrected spectral curve only needs to be input into the modeling model to accurately and efficiently obtain the water quality parameter concentration, making the method of the present invention accurate and efficient.

[0030] 2. In the present invention, the water quality parameters include relatively comprehensive items, which can realize water quality monitoring more comprehensively and accurately.

[0031] 3. In the present invention, when water bodies with different water quality parameter concentrations are selected, the water quality parameter concentrations are selected according to the gradient distribution, which is more efficient and can obtain the reflection spectrum curves corresponding to water bodies with different water quality parameter concentrations more quickly and accurately under the premise of high efficiency.

[0032] 4. The hyperspectral imager for acquiring hyperspectral images in the present invention is mounted on a smart pole, which can realize non-contact water quality monitoring and avoid the shortcomings of traditional contact water environment measurement.

[0033] 5. The present invention also provides a computer program product capable of executing the above method steps, which can promote and apply the method of the present invention and realize integration on corresponding hardware devices. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a schematic diagram of the process of the water quality monitoring method based on hyperspectral multi-parameters of the present invention;

[0035] Figure 2 It is a distribution diagram of the band composition of the hyperspectral imager image in the embodiment of the present invention;

[0036] Figure 3 Schematic diagram of decomposition of a hyperspectral imager image along the row direction in an embodiment of the present invention;

[0037] Figure 4 This is a spectrum curve diagram extracted from a normal water body in an embodiment of the present invention;

[0038] Figure 5 is a spectrum curve diagram extracted from a water body affected by environmental factors in an embodiment of the present invention;

[0039] Figure 6 is a spectrum curve diagram of the external environment in an embodiment of the present invention;

[0040] Figure 7 A schematic diagram of correlation coefficients between 2048 spectral curves proposed for a water body with a turbidity of 13.2 NTU in an embodiment of the present invention and a standard water body curve;

[0041] Figure 8 This is a reflection spectrum curve diagram obtained after calibration with a black and white plate in an embodiment of the present invention;

[0042] Fig. 9 This is a comparison chart of the predicted value and the measured value of the water turbidity predicted by the method of the embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0044] The present invention provides a water quality parameter monitoring method based on hyperspectral multi-parameters, which utilizes hyperspectral technology to realize non-contact measurement of key water quality parameters, improves the influence of factors such as floating objects and bubbles on the water surface on the spectral information of the water body, thereby improving the measurement accuracy of relevant water quality parameters.

[0045] Take a hyperspectral imager mounted on a smart pole as an example. The hyperspectral imager detects spectral information by linear gradient filtering, and the pixel size of the image is 2048*2448. Each image contains 85 bands of wavelength components from 430nm to 880nm. The band composition of the image is as follows: Figure 2 shown.

[0046] When the image acquisition object is a water body, the water body is set to be uniform within the imaging range. According to the imaging characteristics of the image, each row of pixels in the image can be regarded as the spectral curve of this part of the water body. The acquired image can be decomposed into 2048 spectral curves, as shown in the following figure. Figure 3 shown.

[0047] By calculating the correlation coefficients of these 2048 spectral curves with the set standard water body spectral curve, setting a threshold T, and removing the spectral curves with correlation coefficients less than T (these spectral curves may be affected by bubbles or suspended substances, thus changing the spectral characteristics). Calculate the average value of the remaining spectral curves as the final spectral curve of the water body.

[0048] Perform the above processing on both the hyperspectral images of the water bodies participating in the modeling and those to be predicted, to obtain the final spectral curves of the water bodies at different water quality parameter concentrations. Then complete the subsequent modeling and prediction work through machine learning methods.

[0049] As Figure 1 shown, the water quality monitoring method based on hyperspectral multi-parameters of the present invention specifically includes the following steps:

[0050] Step 1, obtain water quality parameters, where the water quality parameters include supplementary chlorophyll a, total nitrogen, total phosphorus, transparency, chemical oxygen demand, turbidity, suspended substances, permanganate index, etc. The method for obtaining water quality parameters can adopt existing traditional methods.

[0051] Step 2, collect the hyperspectral image reflected by the water surface through a hyperspectral imager, and decompose the image into a spectral curves along the row direction, where a is the number of pixel rows in the pixel size of the spectrometer detector;

[0052] Step 3, calculate the correlation coefficient R of these a spectral curves with the set standard water body spectral curve, set a threshold T, and remove the spectral curves with R < T to obtain n spectral curves;

[0053] When calculating the correlation coefficient R, first, it is necessary to preprocess the a spectral curves. Each spectral curve extracts a 1*85-dimensional spectral curve according to the response peak position of each wavelength (there is a wavelength and peak position correspondence table). Then, through the following formula, calculate the correlation coefficient R of the a spectral curves preprocessed by the above steps and the standard water body spectral curve:

[0054]

[0055] where X and Y respectively represent the two spectral curves for calculating the correlation coefficient R, Cov(X,Y) is the covariance of X and Y, and Var[X] and Var[X] are the variances of X and Y respectively.

[0056] Finally, set a threshold T and remove the spectral curves with R < T. The purpose of setting the threshold is to judge whether the spectral curve is affected by bubbles or floating objects on the water surface through the correlation coefficient. The smaller R is, the greater the influence. Therefore, set the threshold T according to the actual calculation situation to remove the spectral curves with small R values.

[0057] Step 4, calculating the average of the n spectral curves as the final spectral curve S of the water body;

[0058] Step 5, perform black and white plate correction on the spectrum curve S to obtain a corrected reflection spectrum curve S';

[0059] Step 6: Process the hyperspectral image data collected from water bodies at different water quality parameter concentrations according to steps 2 to 5 to obtain a modeling data set M = [S′1, S′2, S′3, …, S′ m ];

[0060] Step 7, using a modeling method (PLS, BPNN, etc.) to model the modeling data set M and its corresponding water quality parameter concentration. The specific parameter settings of the modeling method can be adjusted according to actual needs and are not specifically limited;

[0061] Step 8: After the modeling in step S7 is completed, the newly acquired hyperspectral image data can be processed by the model to complete the water quality monitoring. The newly acquired hyperspectral image data is processed through steps 2 to 5, and then input into the model built in step 7 to obtain the water quality parameter concentration corresponding to the newly acquired hyperspectral image data.

[0062] The following is a specific embodiment of the method of the present invention:

[0063] Take the prediction of turbidity concentration in water as an example:

[0064] 1. Collect hyperspectral images of water bodies with different turbidity, such as images of water bodies with a turbidity of 13.2 NTU.

[0065] 2. Decompose the image into 2048 spectral curves along the row direction, pre-process the 2048 spectral curves respectively, and extract a 1*85-dimensional spectral curve according to the response peak position of each wavelength, such as Figure 4 This is the spectrum curve extracted from normal water. Figure 5 The spectral curve extracted for water bodies affected by environmental factors, Figure 6 It is the spectral curve of the external environment.

[0066] 3. Calculate the correlation coefficient between the spectral curve extracted from the hyperspectral image and the standard water body curve extracted. Figures 4 to 6 Taking the corresponding curve as an example, the correlation coefficients between it and the standard water curve are 0.8693, -0.0761, and 0.3484 respectively. The correlation coefficients between the 2048 spectral curves proposed for the water body with turbidity of 13.2 NTU and the standard water curve are as follows Figure 7 shown.

[0067] When setting the threshold T, Figure 5 and Figure 6 This type of curve is eliminated, and in this embodiment, the threshold is set to T = 0.6. After elimination, 817 spectral curves are obtained, and these 817 spectral curves are averaged as the final spectral curve of the water body with turbidity of 13.2 NTU.

[0068] The final spectral curves of 18 turbidity water bodies were obtained through the above method.

[0069] 4. Perform black and white plate correction on the final spectral curve of each turbidity water body calculated to obtain the corrected reflectance spectral curve as shown below: Figure 8 The specific turbidity of each turbidity water body is shown in Table 1.

[0070] Table 1 Water turbidity table

[0071]

[0072]

[0073] 5. The corrected reflectance spectrum curve and its corresponding turbidity are modeled using PLS, and the leave-one-out method (one of the cross-validation methods) is used to verify the modeling effect. When PLS modeling is performed, the number of principal components is set to 2. The predicted values ​​and measured values ​​of the 18 samples are as follows: Fig. 9 As shown, Fig. 9 The small circles in the figure are measured values, and the predicted results R 2 =0.9899,RMSEP=1.6851.

[0074] It can also be seen from the above embodiments that the water quality monitoring method of the present invention has high accuracy and is easy to operate.

[0075] The present invention can also form a computer program product based on the hyperspectral multi-parameter water quality monitoring method, which includes a computer program that implements the steps of the hyperspectral multi-parameter water quality monitoring method when executed by a processor.

[0076] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A water quality monitoring method based on hyperspectral multi-parameters, characterized in that: The following steps are involved: S1, obtain the modeling dataset S1.1, using a hyperspectral imager to collect a hyperspectral image reflected by the water surface, wherein the hyperspectral imager realizes spectral information detection by linear gradient filtering; decomposing the hyperspectral image reflected by the water surface into a spectral curves along the row direction, wherein a is the number of pixel rows of the spectrometer detector; S1.2, respectively calculate the correlation coefficient R between a spectral curve and the standard water spectral curve, and set a threshold T. If R < T, the corresponding spectral curve is eliminated, otherwise, the corresponding spectral curve is retained to obtain n spectral curves, where n is the number of retained spectral curves, n ≤ a; The correlation coefficient R between the a spectral curve and the standard water spectral curve is calculated respectively, specifically: For each spectral curve, a multidimensional spectral curve is extracted according to the response peak position of each wavelength. The dimension of the multidimensional spectral curve is the number of bands of the hyperspectral image. Then, the correlation coefficient R between the spectral curve a and the standard water spectral curve is calculated by the following formula: Where X and Y represent the two spectral curves for calculating the correlation coefficient R, Cov(X,Y) is the covariance of X and Y, Var[X] and Var[Y] are the variances of X and Y respectively; S1.3, calculating the average of n spectral curves, and taking the average as the final spectral curve S of the water body; S1.4, performing black and white plate correction on the spectrum curve S to obtain a corrected reflection spectrum curve S'; S1.5, for water bodies with different water quality parameter concentrations, repeat steps S1.1 to S1.4 to obtain the reflection spectrum curves S' corresponding to the water bodies with m water quality parameter concentrations, forming the modeling data set M: M=[S′1,…,S ′ m ] Where m is an integer greater than 1, S′1 is the reflectance spectrum curve corresponding to the water body with the first water quality parameter concentration, S ′ m is the reflectance spectrum curve corresponding to the water body with the mth water quality parameter concentration; S2, get the modeling model The reflectance spectrum curves in the modeling data set M and the concentrations of water quality parameters corresponding to each reflectance spectrum curve are modeled by partial least squares PLS or back propagation neural network BPNN to obtain a modeling model; S3, monitor water quality The hyperspectral image reflected by the water surface is processed through steps S1.1 to S1.4 and then input into the modeling model to obtain the corresponding water quality parameter concentration.

2. A water quality monitoring method based on hyperspectral multi-parameters as claimed in claim 1, characterized in that: In step S1.5, the water quality parameters include supplemental chlorophyll a, total nitrogen, total phosphorus, transparency, chemical oxygen demand, turbidity, suspended solids and permanganate index.

3. A water quality monitoring method based on hyperspectral multi-parameters as claimed in claim 2, characterized in that: In step S1.5, the water bodies with different water quality parameter concentrations are specifically water bodies with different water quality parameter concentrations distributed in a gradient.

4. A water quality monitoring method based on hyperspectral multi-parameters as claimed in claim 3, characterized in that: The hyperspectral image is specifically acquired by a hyperspectral imager mounted on the smart pole.

5. A computer program product, comprising a computer program, characterized in that: When the program is executed by a processor, the steps of a water quality monitoring method based on hyperspectral multi-parameters described in any one of claims 1 to 4 are implemented.

Images