Monitoring method based on unmanned aerial vehicle for water quality

Abstract

The invention relates to a water quality unmanned aerial vehicle-based monitoring method, which comprises an unmanned aerial vehicle body, a high-definition camera, a hyperspectral sensor, a holder, a processor and a wireless transmission device, and is characterized in that the processor and the wireless transmission device are mounted in the unmanned aerial vehicle body, and the wireless transmission device is used for wirelessly transmitting videos and data processed by the processor; the holder is mounted at the bottom of the unmanned aerial vehicle body, and the high-definition camera and the hyperspectral sensor are mounted on the holder. The system can monitor various water conservancy data such as water flow velocity, water color and water quality data. A flight monitoring task can be executed according to a specified GPS coordinate and a fixed-time river channel monitoring task; and dangerous behaviors around rivers, lakes and reservoirs can be identified and detected, such as illegal pollution discharge, illegal collection and leakage, people fishing, people entering water and the like. The problem of large-area and timely water regimen monitoring is solved, the deployment cost is reduced, and the maintenance efficiency is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of water quality monitoring, and in particular to a water quality monitoring method based on a water body unmanned aerial vehicle. Background Art

[0002] With the development of AI intelligent technology in recent years, it has been able to maturely identify and monitor water levels, flow rates, and flow data of rivers, lakes, and reservoirs; however, it has put forward higher requirements for monitoring water bodies and water quality, quickly grasping water data, and timely warning. In the past, water body and water quality monitoring required the installation of monitoring equipment and surveillance cameras in multiple locations in the river. It was difficult to deploy over a large area, the equipment cost was high, the maintenance cost was high, and the data aggregation system was complex. At present, drone monitoring of rivers only stays at the monitoring of water levels and flow rates, and lacks water quality data monitoring of large areas of rivers.

[0003] With the development of spectral remote sensing technology and UAV flight control, the requirements for rapid and large-area synchronous monitoring have been met; UAV platforms equipped with hyperspectral sensors can obtain remote sensing data with high spatial, high temporal and high spectral resolution, and use mature AI intelligent analysis to achieve long-term and accurate water quality monitoring of large areas and narrow rivers; it is of great significance for the continuous monitoring of water conservancy and water environment and emergency key inspections. Summary of the invention

[0004] The purpose of the present invention is to overcome the shortcomings of the prior art and propose a water quality monitoring method based on a water body drone. By taking advantage of the rapid deployment of drones, it is possible to quickly monitor and collect inverted water quality data over long distances.

[0005] The present invention solves the technical problem by adopting the following technical solutions:

[0006] A water quality monitoring method based on a water body drone, wherein the water quality drone used includes a drone body, a high-definition camera, a hyperspectral sensor, a gimbal, a processor and a wireless transmission device, wherein the processor and the wireless transmission device are installed inside the drone body, the processor is used to process videos and data collected by the high-definition camera and the hyperspectral sensor, the wireless transmission device is used to wirelessly transmit the videos and data processed by the processor, the gimbal is installed at the bottom of the drone body, the high-definition camera and the hyperspectral sensor are mounted on the gimbal, the high-definition camera is used to collect video image information of the water body, and the hyperspectral sensor is used to collect spectral information.

[0007] Moreover, the water quality UAV also includes a diffuse reflection reference board, wherein the diffuse reflection reference board is installed below the gimbal, and the diffuse reflection reference board is used to measure the reflectivity and downlink radiance of the position.

[0008] Furthermore, the following steps are included:

[0009] Step 1: The drone flies to the destination according to the specified mission; first, it identifies the water surface situation from a high altitude, selects the measurement location, and descends to the effective area to perform spectral reflectance measurement;

[0010] Step 2: Obtain the reflectivity R at the measurement position through the diffuse reflection reference plate ref (λ) and the downlink radiance L ref (λ), obtain the total solar irradiance, and synchronously measure the sky light radiance above the reference plate according to the above-water spectral measurement method;

[0011] Step 3: Collect hyperspectral images through a hyperspectral sensor and remove noise;

[0012] Step 4: Calculate the Hueangle of rivers and lakes based on the noise-removed hyperspectral image;

[0013] Step 5: Classify the water bodies according to the Hueangle of rivers and lakes;

[0014] Step 6: Invert the water quality parameters according to the classification to obtain the water body status.

[0015] Moreover, the effective area in step 1 is 1m 2 .

[0016] Moreover, the total solar irradiance E calculated in step 2 is d The specific implementation method of (λ) is:

[0017] E d (λ)=π·L ref (λ) / R ref (λ).

[0018] Moreover, in step 2, the sky light radiance L above the reference plate is synchronously measured according to the spectrum measurement method above the water surface. sky The specific implementation method of (λ) is: the incident radiance of sky light L measured synchronously on the ground sky is the sky light radiance above the drone image, the Fresnel reflectivity r is 0.023, and the drone hyperspectral remote sensing reflectivity for:

[0019]

[0020] The accuracy of remote sensing reflectivity is evaluated by root mean square error (RMSE), mean unbiased absolute percentage error (ε) and Pearson correlation coefficient (R):

[0021]

[0022] Among them, u is the remote sensing reflectivity calculated by the UAV; f is the measured remote sensing reflectivity; and n is the number of verification points.

[0023] Moreover, the specific implementation method of step 3 is: using fast Fourier transform FFT time domain and frequency domain transform to remove the strip noise of the hyperspectral image, and evaluating the effect of strip noise removal by calculating the strip measurement value Si of each column of the hyperspectral image; using the noise-sensitive minimum noise separation method MNF to realize the removal of spectral noise, and using the local mean and local standard deviation method to calculate the signal-to-noise ratio and noise standard deviation of the hyperspectral image to test the denoising effect of the hyperspectral image.

[0024] Moreover, the specific implementation method of step 4 is:

[0025]

[0026] Among them, R rs is the remote sensing reflectance of water body, λi is the characteristic value of spectral wavelength.

[0027] Moreover, the specific implementation method of step 5 is: according to the Hueangle values ​​of different water bodies, the water bodies are divided into green abnormal water bodies, general water bodies and yellow-brown abnormal water bodies.

[0028] Moreover, the specific implementation method of step 6 is: by performing statistical analysis on the synchronously measured remote sensing reflectance f and water quality parameters, selecting the band or band combination with the highest correlation for different water quality parameters as independent variables, constructing an inversion model through exponential, power function, linear, polynomial or partial least squares regression method, and evaluating the inversion effect of the model through R2 and RMSE.

[0029] The advantages and positive effects of the present invention are:

[0030] The present invention includes an unmanned aerial vehicle body, a high-definition camera, a hyperspectral sensor, a pan-tilt, a processor and a wireless transmission device, wherein the processor and the wireless transmission device are installed inside the unmanned aerial vehicle body, the processor is used to process the video and data collected by the high-definition camera and the hyperspectral sensor, the wireless transmission device is used to wirelessly transmit the video and data processed by the processor, the pan-tilt is installed at the bottom of the unmanned aerial vehicle body, the high-definition camera and the hyperspectral sensor are mounted on the pan-tilt, the high-definition camera is used to collect video image information of the water body, and the hyperspectral sensor is used to collect spectral information. The present invention can monitor a variety of water conservancy data: such as water flow rate, water body color, water quality data, etc. It can perform flight monitoring tasks according to specified GPS coordinates and scheduled river monitoring tasks; it can also identify and detect dangerous behaviors around rivers, lakes, and reservoirs, such as illegal sewage discharge, illegal mining and leakage, fishing, and people going into the water. It solves the problem of large-scale and timely monitoring of water conditions, reduces deployment costs, and improves maintenance efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 This is a structural diagram of the drone of the present invention;

[0032] Figure 2 This is a schematic diagram of the noise reduction effect of the present invention;

[0033] Figure 3 It is a schematic diagram of water body differentiation according to the present invention. DETAILED DESCRIPTION

[0034] The present invention is further described in detail below with reference to the accompanying drawings.

[0035] A water quality monitoring method based on drones, such as Figure 1 As shown, the water quality UAV includes a UAV body, a high-definition camera, a hyperspectral sensor, a gimbal, a processor and a wireless transmission device, wherein the processor and the wireless transmission device are installed inside the UAV body, the processor is used to process the video and data collected by the high-definition camera and the hyperspectral sensor, the wireless transmission device is used to wirelessly transmit the video and data processed by the processor, the gimbal is installed at the bottom of the UAV body, the high-definition camera and the hyperspectral sensor are mounted on the gimbal, the high-definition camera is used to collect video image information of the water body, and the hyperspectral sensor is used to collect spectral information.

[0036] It also includes a diffuse reflection reference plate, wherein the diffuse reflection reference plate is installed below the pan / tilt platform, and the diffuse reflection reference plate is used to measure the reflectivity and downward radiance of the position.

[0037] A water quality monitoring method based on a water body drone comprises the following steps:

[0038] Step 1: The drone flies to the destination according to the specified mission; first, it identifies the water surface from a high altitude, selects the measurement location, and descends to the effective area for spectral reflectance measurement. The effective area is 1m 2 .

[0039] Step 2: Obtain the reflectivity R at the measurement position through the diffuse reflection reference plate ref (λ) and the downlink radiance L ref The total solar irradiance is obtained by measuring the spectral measurement method above the water surface, and the sky radiance above the reference plate is measured synchronously.

[0040] Use a diffuse reflection reference plate with a reflectivity of 20% to 30% and obtain the reflectivity R at the measurement position through the diffuse reflection reference plate. ref (λ) and the downlink radiance L ref (λ), calculate the total solar irradiance:

[0041] E d (λ)=π·Lref (λ) / R ref (λ).

[0042] The sky light radiance L above the reference plate is measured synchronously according to the above-water spectral measurement method. sky (λ), the incident radiance of skylight L measured synchronously on the ground sky is the sky light radiance above the drone image, the Fresnel reflectivity r is 0.023, and the drone hyperspectral remote sensing reflectivity for:

[0043]

[0044] The accuracy of remote sensing reflectivity is evaluated by root mean square error (RMSE), mean unbiased absolute percentage error (ε) and Pearson correlation coefficient (R):

[0045]

[0046] Among them, u is the remote sensing reflectivity calculated by the UAV; f is the measured remote sensing reflectivity; and n is the number of verification points.

[0047] Step 3: Collect hyperspectral images through hyperspectral sensors and remove noise.

[0048] Fast Fourier transform (FFT) time-domain and frequency-domain transforms are used to remove the strip noise of hyperspectral images. The strip metric value (Si) of each column (vertical distribution of strip noise) of the hyperspectral image is calculated to evaluate the effect of strip noise removal. The noise-sensitive minimum noise separation method (MNF) is used to remove spectral noise. The signal-to-noise ratio and noise standard deviation of the hyperspectral image are calculated using the local mean and local standard deviation methods to test the denoising effect of the hyperspectral image.

[0049] like Figure 2 (a) shows that after using FFT to remove the bands, the overall Si has been significantly reduced, which is lower than the standard of sufficient band removal Si = 0.005, proving that this method can effectively remove the band noise of hyperspectral data. Figure 2 (b) shows that the MNF method effectively removes a large amount of noise in the spectral dimension and obtains a relatively smooth spectral curve. Figure 2 (c) and Figure 2 (d) shows that after MNF noise removal, the signal-to-noise ratio and noise standard deviation are greatly improved. The mean signal-to-noise ratio is increased by 128%, and the noise standard deviation is also significantly reduced, with an average value of 3.42×10-4 ( Figure 2 (d)); After FFT removed the strip noise, the signal-to-noise ratio and noise standard deviation did not improve significantly.

[0050] Step 4: Calculate the Hueangle of rivers and lakes based on the denoised hyperspectral image.

[0051] The CIE-XYZ color standard system and quantitative color description method developed by the International Commission on Illumination (CIE) are used to calculate the Hueangle of river and lake water bodies based on the UAV hyperspectral data collected in this study. In addition, the dominant wavelength of a given spectrum is also used to represent it. Because AVW includes ultraviolet and near-infrared bands, it has a higher range of variation relative to Hueangle for extreme waters that are bluer or redder in color. Therefore, in order to make full use of the obtained hyperspectral information to accurately quantify the water color of inland water bodies, the present invention uses the full-band hyperspectral data (400-800nm) of the drone to calculate AVW from the perspective of the dominant wavelength of water color. The calculation formula is:

[0052]

[0053] Among them, R rs is the remote sensing reflectance of water body, λi is the characteristic value of spectral wavelength.

[0054] Step 5: Classify water bodies according to the Hueangle of rivers and lakes.

[0055] Refer to the method of setting Hueangle threshold for the classification of inland river and lake water bodies. Figure 3 As shown in the figure, 19 high-spectral reflectance curves of different types of water bodies (5 green water bodies with duckweed floating, 7 general water bodies and 7 yellow-brown water bodies near pollution sources) were selected from the measured data to calculate Hueangle. The water body with duckweed floating is green and has plant spectral characteristics. The main components of the yellow-brown water body near the pollution source are non-pigmented particulate matter or colored dissolved organic matter. The humic acid and fulvic acid contained in the colored dissolved organic matter will make the water body appear yellow-brown as the concentration increases. According to the Hueangle values ​​of the selected different water bodies, the water bodies are divided into three categories: green abnormal water bodies (Hueangle≤218°), general water bodies (218°≤Hueangle≤225°), and yellow-brown abnormal water bodies (Hueangle≥225°)

[0056] Step 6: Invert the water quality parameters according to the classification to obtain the water body status.

[0057] Spectral remote sensing inversion is performed by statistically analyzing the synchronously measured remote sensing reflectance f and the water quality parameters (Chl-a, TSM, CDOM, turbidity, TN, TP) shown in Table 1.

[0058] Table 1 Water quality parameters

[0059]

[0060] The bands or band combinations with the highest correlation of different water quality parameters are selected as independent variables, and the inversion model is constructed by exponential, power function, linear, polynomial, partial least squares regression and other methods, as shown in Table 2. The inversion effect of the model is evaluated by R2 and RMSE. The water quality parameter inversion model is as follows:

[0061] Table 2 Inversion model

[0062]

[0063] The TLI index of rivers and lakes was estimated based on the concentrations of Chl-a, TN and TP, with reference to the calculation method of the comprehensive trophic index (TLI). The nutrient status of water bodies was divided into five states: oligotrophic (TLI < 30), mesotrophic (30 ≤ TLI ≤ 50), slightly eutrophic (50 < TLI ≤ 60), moderately eutrophic (60 < TLI ≤ 70), and severely eutrophic (70 < TLI) by the TLI index of water bodies.

[0064] According to the above-mentioned water quality monitoring method based on a water body drone, the effect of the present invention has been verified by measuring a river channel.

[0065] Step 1: Construct a drone for monitoring water quality according to the structure;

[0066] Step 2: Deploy drones to monitor water quality and perform flight monitoring tasks according to scheduled or temporary tasks;

[0067] Step 3: Fly to the designated location to automatically monitor and detect water color and water quality data, and transmit the inverted data back to the command platform in real time through 4G or image transmission channels to summarize the data and conduct early warning processing to achieve overall water quality monitoring.

[0068] The present invention makes full use of the characteristics of UAV flight, such as fast, accurate, stable and convenient, to monitor the water quality of rivers and lakes over large areas or long distances. At the same time, the spectral remote sensing data is quickly collected, filtered and denoised, and the water color and water quality are inverted to output high-precision data. The present invention can monitor a variety of water conservancy data: such as water flow rate, water color, water quality data, etc. It can perform flight monitoring tasks according to specified GPS coordinates and scheduled river monitoring tasks; it can also identify and detect dangerous behaviors around rivers, lakes and reservoirs, such as illegal sewage discharge, illegal mining and leakage, fishing, and people going into the water. It solves the problem of large-scale and timely monitoring of water conditions, reduces deployment costs, and improves maintenance efficiency.

[0069] It should be emphasized that the embodiments described in the present invention are illustrative rather than restrictive. Therefore, the present invention includes but is not limited to the embodiments described in the specific implementation manner. Any other implementation manners derived by those skilled in the art based on the technical solution of the present invention also fall within the scope of protection of the present invention.

Claims

1. A water quality monitoring method based on a water quality drone, characterized by: The water quality drone used includes a drone body, a high-definition camera, a hyperspectral sensor, a gimbal, a processor and a wireless transmission device, wherein the processor and the wireless transmission device are installed inside the drone body, the processor is used to process the video and data collected by the high-definition camera and the hyperspectral sensor, the wireless transmission device is used to wirelessly transmit the video and data processed by the processor, the gimbal is installed at the bottom of the drone body, the high-definition camera and the hyperspectral sensor are mounted on the gimbal, the high-definition camera is used to collect video image information of the water body, and the hyperspectral sensor is used to collect spectral information.

2. The water quality monitoring method based on a water quality drone according to claim 1 is characterized in that: The water quality UAV also includes a diffuse reflection reference board, wherein the diffuse reflection reference board is installed below the gimbal, and the diffuse reflection reference board is used to measure the reflectivity and downlink radiance of the position.

3. The water quality monitoring method based on a water quality drone according to claim 1 is characterized in that: The following steps are involved: Step 1: The drone flies to the destination according to the specified mission; first, it identifies the water surface situation from a high altitude, selects the measurement location, and descends to the effective area to perform spectral reflectance measurement; Step 2: Obtain the reflectivity R at the measurement position through the diffuse reflection reference plate ref (λ) and the downlink radiance L ref (λ), obtain the total solar irradiance, and synchronously measure the sky light radiance above the reference plate according to the above-water spectral measurement method; Step 3: Collect hyperspectral images through a hyperspectral sensor and remove noise; Step 4: Calculate the Hueangle of rivers and lakes based on the noise-removed hyperspectral image; Step 5: Classify the water bodies according to the Hueangle of rivers and lakes; Step 6: Invert the water quality parameters according to the classification to obtain the water body status.

4. The water quality monitoring method based on a water quality drone according to claim 3 is characterized by: The effective area in step 1 is 1m 2 .

5. The monitoring method of a water quality monitoring method based on a water quality drone according to claim 3 is characterized in that: The total solar irradiance E is calculated in step 2 d The specific implementation method of (λ) is: E d (λ)=π·L ref (L) / R ref (l)。 6. The method for monitoring water quality using a water quality drone according to claim 3, characterized in that: In step 2, the sky light radiance L above the reference plate is synchronously measured according to the spectrum measurement method above the water surface. sky The specific implementation method of (λ) is: the incident radiance of sky light L measured synchronously on the ground sky is the sky light radiance above the drone image, the Fresnel reflectivity r is 0.023, and the drone hyperspectral remote sensing reflectivity for: The accuracy of remote sensing reflectivity is evaluated by root mean square error (RMSE), mean unbiased absolute percentage error (ε) and Pearson correlation coefficient (R): Among them, u is the remote sensing reflectivity calculated by the UAV; f is the measured remote sensing reflectivity; and n is the number of verification points.

7. The method for monitoring water quality using a drone according to claim 3, characterized in that: The specific implementation method of step 3 is: use fast Fourier transform FFT time domain and frequency domain transform to remove the strip noise of the hyperspectral image, and evaluate the effect of strip noise removal by calculating the strip measurement value Si of each column of the hyperspectral image; use the noise-sensitive minimum noise separation method MNF to realize the removal of spectral noise, use the local mean and local standard deviation method to calculate the signal-to-noise ratio and noise standard deviation of the hyperspectral image, and test the denoising effect of the hyperspectral image.

8. The method for monitoring water quality using a water quality drone according to claim 3, characterized in that: The specific implementation method of step 4 is: Among them, R rs is the remote sensing reflectance of water body, λi is the characteristic value of spectral wavelength.

9. The method for monitoring water quality using a drone according to claim 3, characterized in that: The specific implementation method of step 5 is: according to the Hueangle values ​​of different water bodies, the water bodies are divided into green abnormal water bodies, general water bodies and yellow-brown abnormal water bodies.

10. The method for monitoring water quality using a water quality drone according to claim 3, characterized in that: The specific implementation method of step 6 is: through statistical analysis of the synchronously measured remote sensing reflectance f and water quality parameters, the band or band combination with the highest correlation of different water quality parameters is selected as the independent variable, the inversion model is constructed through the exponential, power function, linear, polynomial or partial least squares regression method, and the inversion effect of the model is evaluated through R2 and RMSE.

Images