Water body nitrate nitrogen concentration high-frequency monitoring method and system based on water surface hyperspectral data

Abstract

The invention discloses a water body nitrate concentration high-frequency monitoring method and system based on water surface hyperspectral data, and the method comprises the following steps: arranging a hyperspectrometer on the surface of a monitored water body for continuous monitoring, and obtaining water body spectral absorbance data; regularly collecting a water sample on site, and measuring the nitrate nitrogen concentration of the water sample; matching the spectral absorbance data with the nitrate nitrogen data, and establishing a sample data set; training a PLS model by using the data set, and constructing a nitrate nitrogen inversion model; and the established inversion model is combined with a hyperspectrometer to realize automatic long-time high-frequency monitoring of nitrate nitrogen.

Description

Technical Field

[0001] The present invention relates to the technical field of hyperspectral remote sensing monitoring, and in particular to a method and system for high-frequency monitoring of nitrate-nitrogen concentration in water bodies based on water surface hyperspectral data. Background Art

[0002] The test results of water quality parameters are an important basis for evaluating water quality and carrying out water pollution prevention and control work, especially for surface water, such as rivers, lakes and reservoirs. Therefore, convenient and fast water quality testing is very necessary. Seeking a method and system for rapid and synchronous testing of water quality parameters that can be implemented on a large scale is conducive to the protection of the current water environment and the control of water pollution, and is also a necessary work to promote sustainable social and economic development.

[0003] Nitrate nitrogen is a water quality indicator that effectively reflects the concentration of nitrates in water bodies and can be used to measure the degree of pollution in water bodies. In aerobic water bodies, bound nitrogen has a tendency to convert into nitrate nitrogen, so nitrate nitrogen concentration is an important water quality indicator for monitoring nitrogen elements. Excessive nitrate nitrogen content in water bodies will cause drinking water safety problems, so the significance of nitrate nitrogen monitoring is self-evident.

[0004] At present, monitoring of nitrate nitrogen in water bodies can only use traditional chemical analysis methods, which takes a huge amount of time and financial resources and cannot achieve rapid and real-time monitoring. Summary of the invention

[0005] The present invention provides a high-frequency monitoring method and system for nitrate-nitrogen concentration in water bodies based on water surface hyperspectral data, aiming to solve the problem that nitrate-nitrogen monitoring in existing water bodies can only be carried out using traditional chemical analysis methods, which costs a lot of time and financial resources and cannot achieve rapid real-time monitoring.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is: The high-frequency monitoring method of water body nitrate nitrogen concentration based on water surface hyperspectral data includes the following steps: S1. Setting a hyperspectral spectrometer on the surface of the monitored water body for continuous monitoring to obtain the spectral absorbance data of the water body; S2. Collect water samples regularly on site and measure the nitrate nitrogen concentration of the water samples; S3, matching the spectral absorbance data with the nitrate nitrogen data to establish a sample data set; S4. Use the data set to train the PLS model and build a nitrate-nitrogen inversion model; S5. Combine the established inversion model with the hyperspectral instrument to realize automatic long-term and high-frequency monitoring of nitrate nitrogen.

[0007] Preferably, the continuous monitoring in S1 comprises the following steps: using an active light source of a hyperspectrometer to emit light and reflect and receive light, measuring the brightness values ​​of the emitted and received light, and calculating the spectral absorbance.

[0008] As a more preferred embodiment, in the continuous monitoring of S1, the monitoring frequency of the hyperspectrometer is at most 1 minute / time, and the spectral detector of the hyperspectrometer is 20 cm below the water surface.

[0009] Furthermore, in the continuous monitoring of S1, the spectral absorbance obtained by the hyperspectrometer in each monitoring is taken at intervals of 2 nm within the range of 201-1000 nm, and a total of 405 band values ​​are selected as the spectral absorbance.

[0010] Furthermore, in S2, the nitrate nitrogen concentration of the water sample is determined by ultraviolet spectrophotometry.

[0011] Specifically, S4 also includes the following steps: preprocessing the spectral absorbance data to eliminate the influence of abnormal values ​​and jump values ​​on the spectral absorbance data, and then performing SG filtering, and then calculating the first-order derivative of the result to obtain the preprocessed spectral absorbance data.

[0012] More specifically, the pre-processed spectral absorbance data are screened for sensitive bands corresponding to water quality indicators, and the spectral absorbance data of 200-400nm are selected as the sensitive band for inverting nitrate nitrogen.

[0013] In detail, in S4, an inversion model is constructed for the pre-processed spectral absorbance data based on the PLS method, the pre-processed spectral absorbance data is used as the model input element, and the corresponding matched nitrate nitrogen data is used as the model output element.

[0014] In more detail, in S4, the number of principal components of the PLS model is set to 4, the number of random seeds is set to 40, and the nitrate-nitrogen inversion model is established.

[0015] A high-frequency monitoring system for nitrate-nitrogen concentration in water based on water surface hyperspectral data is used to perform high-frequency monitoring of nitrate-nitrogen concentration in water based on the above-mentioned high-frequency monitoring method for nitrate-nitrogen concentration in water based on water surface hyperspectral data, including a monitoring module, a wireless communication module and a processing module, wherein the monitoring module includes an underwater water quality spectrometer for floating on the water surface to monitor the spectral absorbance data of water in real time; The wireless communication module includes a wireless communication device, and the water spectral absorbance data monitored by the monitoring module is transmitted to the pan-tilt station in real time through the wireless communication device; The processing module includes a computer, which obtains water body spectral absorbance data from the pan-tilt platform through a wireless communication module, and outputs corresponding matching nitrate nitrogen data according to an established inversion model.

[0016] Beneficial effects of the present invention: 1. The method is innovative in method, which can realize the automatic monitoring method of nitrate nitrogen using spectral inversion, and on this basis, it can carry out long-term continuous high-frequency monitoring, up to the minute level, and is applicable in different working conditions; 2. Compared with other spectral monitoring methods, the present invention uses hyperspectral monitoring, which has nanometer-level spectral resolution, high sensitivity, and can provide more water body information; 3. Compared with the traditional remote sensing monitoring method, the hyperspectral instrument of this method monitors directly in the water body, which will not be affected by environmental factors such as clouds, fog and atmosphere, avoids the complex pre-processing process, and can effectively improve the monitoring accuracy; 4. The hyperspectrometer used in the present invention uses active light source monitoring. The light intensity is the same each time monitoring, and the monitored spectrum will not be different due to the different daily natural light intensity. This method can realize continuous monitoring at night when there is no light source or in rainy weather with reduced light source. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a flow chart of the automatic nitric nitrogen monitoring method based on hyperspectral data of the present invention; Figure 2 This is a comparison chart of the construction and verification of the nitrate-nitrogen partial least squares model in the implementation case of the present invention, and from left to right are the model verification results based on the training set, verification set and all data respectively. DETAILED DESCRIPTION

[0018] As follows, embodiments are further described with reference to the accompanying drawings.

[0019] like Figure 1 As shown, as a preferred embodiment 1, a high-frequency monitoring method for nitrate-nitrogen concentration in water based on water surface hyperspectral data includes the following steps: S1. Setting a hyperspectral spectrometer on the surface of the monitored water body for continuous monitoring to obtain the spectral absorbance data of the water body; S2. Collect water samples regularly on site and measure the nitrate nitrogen concentration of the water samples; S3, matching the spectral absorbance data with the nitrate nitrogen data to establish a sample data set; S4. Use the data set to train the PLS model and build a nitrate-nitrogen inversion model; S5. Combine the established inversion model with the hyperspectral instrument to realize automatic long-term and high-frequency monitoring of nitrate nitrogen.

[0020] Preferably, the continuous monitoring in S1 comprises the following steps: using an active light source of a hyperspectrometer to emit light and reflect and receive light, measuring the brightness values ​​of the emitted and received light, and calculating the spectral absorbance.

[0021] As a more preferred embodiment, in the continuous monitoring of S1, the monitoring frequency of the hyperspectrometer is at most 1 minute / time, and the spectral detector of the hyperspectrometer is 20 cm below the water surface.

[0022] Furthermore, in the continuous monitoring of S1, the spectral absorbance obtained by the hyperspectrometer in each monitoring is taken at intervals of 2 nm within the range of 201-1000 nm, and a total of 405 band values ​​are selected as the spectral absorbance.

[0023] Furthermore, in S2, the nitrate nitrogen concentration of the water sample is determined by ultraviolet spectrophotometry.

[0024] Specifically, S4 also includes the following steps: preprocessing the spectral absorbance data to eliminate the influence of outliers and jump values ​​on the spectral absorbance data, and then performing SG filtering, after which the first-order derivative of the result is calculated to obtain the preprocessed spectral absorbance data to highlight the spectral features and further improve the model inversion accuracy.

[0025] More specifically, the pre-processed spectral absorbance data are screened for sensitive bands corresponding to water quality indicators, and the spectral absorbance data of 200-400nm are selected as the sensitive band for inverting nitrate nitrogen.

[0026] In detail, in S4, an inversion model is constructed for the pre-processed spectral absorbance data based on the PLS method, the pre-processed spectral absorbance data is used as the model input element, and the corresponding matched nitrate nitrogen data is used as the model output element.

[0027] In more detail, in S4, the number of principal components of the PLS model is set to 4, the number of random seeds is set to 40, and the nitrate-nitrogen inversion model is established.

[0028] As a preferred embodiment 2, a high-frequency monitoring method for nitrate-nitrogen concentration in water based on water surface hyperspectral data comprises the following steps: S1. Place a hyperspectrometer in the river; the hyperspectrometer is linked to the buoy, and the spectral detector of the hyperspectrometer is 20cm below the water surface. The active light source of the hyperspectrometer emits light and reflects and receives it, measures the brightness (DN) value of the emitted and received light, and calculates the spectral absorbance. The observation results are stored in the background system in real time. Carry out continuous monitoring under different working conditions (rain, cloudy, wind and waves, sunny and algal bloom outbreaks), with a daily monitoring time of 9:00-18:00 and a monitoring frequency of 1h / time; S2. Collect water from 1 cm below the water surface on site, store it in cold storage, and send it back to the laboratory regularly. Perform experimental analysis using the ultraviolet spectrophotometer method in the Environmental Protection Industry Standard of the People's Republic of China (HJ / T 346-2007) to obtain nitrate nitrogen data. S3, matching spectral absorbance data and nitrate nitrogen data to construct a sample data set; S4. Use the data set to train the partial least squares (PLS) model and construct a nitrate-nitrogen inversion model.

[0029] Data preprocessing: Remove the influence of outliers and jump values ​​on the spectral absorbance data, and then perform SG filtering. After that, calculate the first-order derivative of the result to obtain the preprocessed spectral absorbance data to highlight the spectral characteristics and further improve the model inversion accuracy.

[0030] The pre-processed spectral absorbance data were screened for sensitive bands corresponding to water quality indicators. For nitrate nitrogen, the nitrate nitrogen concentration and spectral absorbance data were plotted, and it was found that the spectral absorbance data in the range of 200-400nm had a significant correlation with the nitrate nitrogen concentration data, so the spectral absorbance data of 200-400nm was selected as the sensitive band for inverting nitrate nitrogen.

[0031] The preprocessed spectral absorbance data is used as the input element, and the nitrate nitrogen concentration is used as the output element. There are 110 sets of data in total, 66 sets are randomly selected as training set data, and 44 sets are test set data.

[0032] Run the PLS model in Python. PLS is a method of many-to-many linear regression modeling, which is suitable for data with many variables and multiple correlations between variables. The inversion model is constructed with the spectral absorbance of the training set as the independent variable and the corresponding nitrate nitrogen concentration as the dependent variable. The parameter settings of PLS ​​are: the number of principal components is 4, and the number of random seeds is 40.

[0033] After the model training is completed, the spectral absorbance data of the test set is brought into the PLS model calculation, and the predicted value is compared with the corresponding measured value to calculate the following evaluation: the index determination coefficient (R 2 ), root mean square error (RMSE), mean absolute error (MAE), relative root mean square error (rRMSE), and mean relative error (MRE).

[0034] The coefficient of determination (R2) is an indicator that reflects the degree of fit between data. 2 It is between 0 and 1, and the closer it is to 1, the better the model fit.

[0035] The root mean square error (RMSE) is used to reflect the quality of the prediction results. It is the square root of the average of the sum of the squares of the differences between all predicted values ​​and measured values. The smaller the RMSE, the more accurate the prediction results.

[0036] Relative root mean square error (rRMSE) is used to reflect the prediction accuracy of the model. It is the RMSE divided by the mean of the measured values. The smaller the rRMSE, the higher the model accuracy.

[0037] The mean absolute error (MAE) is used to reflect the degree of deviation between the predicted value and the measured value. It is the average of the absolute values ​​of the differences between all predicted values ​​and the measured values. The smaller the MAE, the more accurate the prediction result.

[0038] The mean relative error (MRE) is used to reflect the error between the predicted value and the measured value. The absolute value of the difference between all predicted values ​​and the measured value is divided by the measured value and the average is taken. The smaller the MRE, the smaller the error of the predicted value relative to the measured value.

[0039] Finally, according to the evaluation index, the nitrate-nitrogen inversion model of the present invention can achieve very high inspection accuracy.

[0040] like Figure 2 As shown in the figure, for all training set and test set samples, the determination coefficient is as high as 0.90, the root square error is 0.08, the mean absolute error is 0.07, the relative root mean square error is 10.80%, and the average relative error is 9.35%; whether it is a training set or a test set, its evaluation indicators reflect that it has excellent inversion accuracy.

[0041] As a preferred embodiment 3, a high-frequency monitoring system for nitrate-nitrogen concentration in water based on water surface hyperspectral data is used to perform high-frequency monitoring of nitrate-nitrogen concentration in water based on the above-mentioned high-frequency monitoring method for nitrate-nitrogen concentration in water based on water surface hyperspectral data, including a monitoring module, a wireless communication module and a processing module, wherein the monitoring module includes an underwater water quality spectrometer for floating on the water surface to monitor the spectral absorbance data of water in real time; The wireless communication module includes a wireless communication device, and the water spectral absorbance data monitored by the monitoring module is transmitted to the pan-tilt station in real time through the wireless communication device; The processing module includes a computer, which obtains water body spectral absorbance data from the pan-tilt platform through a wireless communication module, and outputs corresponding matching nitrate nitrogen data according to an established inversion model.

[0042] As a more preferred embodiment 4, the underwater water quality spectrometer of the present invention includes a buoy, which floats on the surface of the monitored water body and is fixedly towed to the shore by a towing rope.

[0043] Furthermore, a solar power supply device is provided above the buoy, and a high-spectrometer test water tank is provided below the buoy.

[0044] Furthermore, the water tank is black and light-shielding, and has small holes on the tank wall for water to penetrate and isolate floating objects in the water and reduce interference of external light sources on the spectrometer. The spectral detector of the underwater water quality spectrometer extends into the tank through the buoy.

[0045] Among them, as a more preferred embodiment 5, the underwater water quality spectrometer can use the S2UVL-0003 underwater water quality spectrometer produced by Tianjin Zhongke Puguang.

Claims

1. A high-frequency monitoring method for nitrate-nitrogen concentration in water based on water surface hyperspectral data, characterized in that: The following steps are involved: S1. Setting a hyperspectral spectrometer on the surface of the monitored water body for continuous monitoring to obtain the spectral absorbance data of the water body; S2. Collect water samples regularly on site and measure the nitrate nitrogen concentration of the water samples; S3, matching the spectral absorbance data with the nitrate nitrogen data to establish a sample data set; S4. Use the data set to train the PLS model and build a nitrate-nitrogen inversion model; S5. Combine the established inversion model with the hyperspectral instrument to realize automatic long-term and high-frequency monitoring of nitrate nitrogen.

2. The high-frequency monitoring method for nitrate-nitrogen concentration in water based on water surface hyperspectral data according to claim 1 is characterized in that: The continuous monitoring in S1 includes the following steps: using the active light source of the hyperspectrometer to emit light and reflect and receive it, measuring the brightness values ​​of the emitted and received light, and calculating the spectral absorbance.

3. The high-frequency monitoring method for nitrate-nitrogen concentration in water based on water surface hyperspectral data according to claim 2 is characterized in that: In the continuous monitoring of S1, the monitoring frequency of the hyperspectrometer is at most 1 minute / time, and the spectrum detector of the hyperspectrometer is 20 cm below the water surface.

4. The high-frequency monitoring method for nitrate-nitrogen concentration in water based on water surface hyperspectral data according to claim 3 is characterized in that: In the continuous monitoring of S1, the spectral absorbance obtained by the hyperspectrometer in each monitoring is taken at intervals of 2 nm within the range of 201-1000 nm, and a total of 405 band values ​​are selected as the spectral absorbance.

5. The high-frequency monitoring method for nitrate-nitrogen concentration in water based on water surface hyperspectral data according to claim 4 is characterized in that: In S2, the nitrate nitrogen concentration of the water sample is measured by ultraviolet spectrophotometry.

6. The high-frequency monitoring method for nitrate-nitrogen concentration in water based on water surface hyperspectral data according to claim 5 is characterized in that: The S4 also includes the following steps: preprocessing the spectral absorbance data to eliminate the influence of abnormal values ​​and jump values ​​on the spectral absorbance data, and then performing SG filtering, after which the first-order derivative of the result is calculated to obtain the preprocessed spectral absorbance data.

7. The high-frequency monitoring method for nitrate-nitrogen concentration in water based on water surface hyperspectral data according to claim 6 is characterized in that: The pre-processed spectral absorbance data were used to screen the sensitive bands corresponding to the water quality indicators, and the spectral absorbance data of 200-400nm were selected as the sensitive bands for inverting nitrate nitrogen.

8. The high-frequency monitoring method for nitrate-nitrogen concentration in water based on water surface hyperspectral data according to claim 7 is characterized in that: In the S4, an inversion model is constructed for the pre-processed spectral absorbance data based on the PLS method, the pre-processed spectral absorbance data is used as the model input element, and the corresponding matched nitric nitrogen data is used as the model output element.

9. The high-frequency monitoring method for nitrate-nitrogen concentration in water based on water surface hyperspectral data according to claim 8, characterized in that: In S4, the number of principal components of the PLS model is set to 4, the number of random seeds is set to 40, and a nitrate-nitrogen inversion model is established.

10. A high-frequency monitoring system for nitrate-nitrogen concentration in water based on water surface hyperspectral data, used for high-frequency monitoring of nitrate-nitrogen concentration in water based on water surface hyperspectral data according to any one of claims 1 to 9, characterized in that: It includes a monitoring module, a wireless communication module and a processing module. The monitoring module includes an underwater water quality spectrometer for floating on the water surface to monitor the spectral absorbance data of the water body in real time. The wireless communication module includes a wireless communication device, and the water spectral absorbance data monitored by the monitoring module is transmitted to the pan-tilt station in real time through the wireless communication device; The processing module includes a computer, which obtains water body spectral absorbance data from the pan-tilt platform through a wireless communication module, and outputs corresponding matching nitrate nitrogen data according to an established inversion model.

Images