KR102450019B1 - Water Quality Monitoring Method and System for Using Unmanned Aerial Vehicle - Google Patents

Abstract

A method and system for monitoring water quality using a drone are presented. The water quality monitoring method performed by the water quality monitoring system using a drone implemented through a computer device according to an embodiment includes the steps of mounting a multi-spectral sensor on the drone, and acquiring a multi-spectral image for an area requiring monitoring ; And after converting the multi-spectral image to a vegetation index, the multi-spectral image and the correlation between the vegetation index and water quality factors may include predicting the eutrophication state and water pollution level for each point.
[National R&D project supporting this invention]
[Business name] Ministry of Environment
[Name of project management (specialized) organization] Nakdong River System Management Committee
[Research project name] Environmental basic research project
[Research Title] Investigation of non-point pollution sources such as improvement of watershed environment survey method using drone and loading of yard compost
[Name of project execution institution] Changwon University Industry-Academic Cooperation Foundation
[Research Period] 2019.01.07~2019.12.06

Description

Water Quality Monitoring Method and System for Using Unmanned Aerial Vehicle}

The following examples relate to a method and system for monitoring water quality using a drone, and more specifically, to a method and system for monitoring water quality using a drone that identifies factors causing algal blooms using various image data acquired from a drone it's about

In recent years, eutrophication occurs frequently in Korea due to the ease of inflow of nutrients into the water system due to climate change, and the occurrence of green algae is increasing. When algae occur, it destroys the aquatic ecosystem and directly affects the safety of water resources such as water sources. Currently, as a technology for monitoring algal blooms, on-site sampling-based research that collects and analyzes water directly from the site, and remote sensing research using satellite and aerial images are in progress.

However, in the case of on-site sampling analysis, it is difficult to monitor a wide area, and in the case of satellite images and aerial images, it is difficult to obtain images at a desired time due to low resolution, and it is difficult to understand images of multi-spectral bands, water temperature, and water flow. have.

In order to solve this problem, recently, remote sensing research using unmanned aerial vehicles (UAVs, drones) is being conducted. Since a drone can acquire a desired type of image in a desired area at a desired point in time at a low altitude, research is being conducted to analyze the algae phenomenon using these advantages. However, current research on algal monitoring using drones is limited to classifying areas where algal blooms need to be removed by identifying whether algal blooms have occurred.

Korea Patent No. 10-1863123 relates to a river algae map creation system using such an autonomous driving unmanned aerial vehicle and a movable unmanned floating body. In addition, it describes the technology for a river green algae mapping system using an autonomous driving unmanned aerial vehicle and a mobile unmanned floating vehicle that profiling water quality by water depth.

Korean Patent Registration No. 10-1863123

The embodiments describe a water quality monitoring method and system using a drone, and more specifically, predict the occurrence of algae in advance and effectively prepare for it by identifying factors causing the algal phenomenon before the occurrence of algae using various image data acquired from the drone. provides the skills to

Examples are the nutrient factors that cause the biggest cause of the occurrence of algae and chlorophyll-a, an indirect indicator, through correlation with the multi-spectral image of drones, and the possibility of predicting the occurrence of algae according to the identified nutrient factors. An object of the present invention is to provide a method and system for monitoring water quality using a drone.

In addition, the embodiments determine the water temperature that affects the occurrence of algae through the thermal infrared image of the drone as well as the nutrient factor, and the flow of the water through the image of the drone, a model that more precisely predicts the possibility of the occurrence of algae. An object of the present invention is to provide a method and system for monitoring water quality using a drone that can be provided.

The water quality monitoring method performed by the water quality monitoring system using a drone implemented through a computer device according to an embodiment includes the steps of mounting a multi-spectral sensor on the drone, and acquiring a multi-spectral image for an area requiring monitoring ; And after converting the multi-spectral image to a vegetation index, the multi-spectral image and the correlation between the vegetation index and water quality factors may include predicting the eutrophication state and water pollution level for each point.

The method may further include acquiring a thermal infrared image and an RGB image through the thermal infrared camera and the RGB camera of the drone.

Using the predicted eutrophication state and water pollution level for each point, the water temperature distribution identified through the thermal infrared image, and the water temperature flow identified through the RGB image, the step of identifying the water quality status and the possibility of algae occurrence before the occurrence of algae may include more.

In the step of predicting the eutrophication state and water pollution level for each point through the correlation, the correlation between the multi-spectral image and the vegetation index and the water quality factor can be analyzed after making a water quality prediction model through correlation analysis and regression analysis. .

In the step of predicting the eutrophication state and water pollution level by point through the correlation, chlorophyll-a and nutrient factors contributing to water pollution by inputting vegetation indices to the water quality prediction model produced through correlation analysis and regression analysis can figure out

The vegetation index may be obtained as an image, and the method may further include outputting a water quality distribution map by inputting the vegetation index image, the thermal infrared image, and the RGB image to a neural network-based water quality prediction model.

In the acquiring of the multi-spectral image, the acquired multi-spectral image may be produced as a multi-spectral orthogonal image used as a map of a region by matching images through image processing.

The method may further include the step of identifying the water quality factors contributing to water pollution by performing water quality analysis by collecting water for each region.

A water quality monitoring system using a drone according to another embodiment includes: a multi-spectral image collecting unit that mounts a multi-spectral sensor on the drone and acquires a multi-spectral image for an area requiring monitoring; And after converting the multi-spectral image to the vegetation index, the multi-spectral image and the correlation between the vegetation index and water quality factors, the water pollution degree prediction unit for predicting the eutrophication state and water pollution level for each point can be made including.

It may further include a thermal infrared image and RGB image collecting unit for acquiring the thermal infrared image and the RGB image through the thermal infrared camera and the RGB camera of the drone.

Using the predicted eutrophication state and water pollution level for each point, the water temperature distribution identified through the thermal infrared image, and the water temperature flow identified through the RGB image, the water quality status and the possibility of algal bloom before the occurrence of algal blooms. It may further include a determination unit.

The water pollution level prediction unit may analyze the multi-spectral image and the correlation between the vegetation index and the water quality factor after producing a water quality prediction model through correlation analysis and regression analysis.

The water pollution level prediction unit can identify chlorophyll-a and nutrient factors contributing to water pollution by inputting vegetation indices into the water quality prediction model produced through correlation analysis and regression analysis.

The vegetation index is obtained as an image, and may further include a water quality distribution map providing unit for outputting a water quality distribution map by inputting the vegetation index image, the thermal infrared image, and the RGB image to a neural network-based water quality prediction model.

The multi-spectral image collecting unit, the acquired multi-spectral image may be produced as a multi-spectral orthogonal image used as a map of a region by matching the images through image processing.

According to embodiments, when monitoring algal blooms, it is possible to provide a method and system for monitoring water quality using a drone that can prevent algal blooms in advance by warning in advance of the possibility of algal blooms before occurrence of algal blooms and responding according to the monitoring results. have.

In addition, according to embodiments, it is possible to identify nutrient factors that cause algal blooms in a wide area through drone images without on-site sampling, so monitoring the water quality using a drone that can greatly reduce the time and resources required for monitoring Methods and systems can be provided.

In addition, according to embodiments, it is possible to provide a water quality monitoring method and system using a drone capable of accurately predicting the possibility of algal bloom by identifying the water temperature status and water flow for each area as well as nutrient factors.

1 is a diagram for explaining a process of analyzing a correlation between a multispectral image and a water quality factor according to an embodiment.
2 is a block diagram illustrating a water quality monitoring system using a drone according to an embodiment.
3 is a flowchart illustrating a water quality monitoring method using a drone according to an embodiment.
4 is a diagram illustrating a vegetation index image obtained by converting a multi-spectral image taken by a drone according to an embodiment.
5 is a diagram illustrating a result of producing a water quality prediction model according to a regression analysis according to an embodiment.
6 is a diagram illustrating a result of manufacturing a water quality distribution map according to an exemplary embodiment.

Hereinafter, embodiments will be described with reference to the accompanying drawings. However, the described embodiments may be modified in various other forms, and the scope of the present invention is not limited by the embodiments described below. In addition, various embodiments are provided in order to more completely explain the present invention to those of ordinary skill in the art. The shapes and sizes of elements in the drawings may be exaggerated for clearer description.

The damage to algae is small if it is dealt with in advance, but if, as in previous studies, the algae phenomenon is identified after the occurrence of algae by comparing the image of the factor indicating the algal phenomenon with the image of the drone, the damage would have already occurred and the algae should be removed. It also consumes more resources to respond.

The following embodiments attempt to predict and effectively prepare for the occurrence of algal blooms by identifying factors causing algal blooms before the occurrence of algal blooms by using various image data acquired from drones.

Specifically, the Examples identify the nutrient factor, which is the biggest cause of the occurrence of algae, and chlorophyll-a, an indirect indicator, through correlation with the multi-spectral image of the drone, and determine the possibility of the occurrence of algae according to the identified nutrient factor. predictable. In addition, the embodiments provide a model for more precisely predicting the possibility of algae occurrence by identifying not only nutrient factors but also the water temperature that affects the occurrence of algae through the thermal infrared image of the drone, and the flow of the water through the image of the drone. can do.

1 is a diagram for explaining a process of analyzing a correlation between a multispectral image and a water quality factor according to an embodiment.

Referring to FIG. 1 , the process of this embodiment includes data acquisition through on-site measurement and basic data collection through pre-processing (101 to 107), accuracy verification and correction of measured data (108, 109), and small rivers through analysis was carried out in the process of identifying and evaluating the distribution of water quality (110~112).

For accurate comparative analysis of the collected data, UAV imaging (101), terrestrial spectrometer measurement (102), and collection (103) were performed for two periods on August 9 and September 30. In the case of UAV operation, a flight point was selected through a preliminary on-site investigation, and approval for filming and flight was obtained before proceeding. In addition, GCP (Ground Control Point) was acquired by using RTK equipment in advance to perform geometric correction when producing orthographic images.

During flight, flight altitude, time, area, speed, etc. were planned by using the QgroundControl application in consideration of the specifications of the UAV and the mounted Red edge-M multi-spectral sensor. The images acquired after flight are processed (104) by image pre-processing, initial processing and geometric correction, point cloud generation, radiation correction, and final processing in PIX4D mapper software to obtain Blue, Green, Red, Red edge, NIR (Near-infrared ray) ) Orthoimages of 5 wavelengths were produced. GCPs obtained in advance were used for geometric correction.

Ground spectrometer measurement (102) measures 5 types of land cover and water quality sampling points (105, 106) for rivers, vegetation, roads, yard composts, and plastic houses in order to understand the spectral characteristics of representative materials existing in the research site. carried out.

The measured data was used in the following process to understand the water quality distribution of small rivers, which is the purpose of this example. In the verification and correction process of the UAV image, a regression model suitable for multi-spectral image correction was made based on the ground spectrometer data for each land cover type measured at the same time for the produced orthographic image, and the image was verified and corrected (109). The zonal statistics tool of ArcGIS 10.6 software was used to extract the image value of the produced orthogonal image, and the terrestrial spectrometer value was obtained from the overlapping part with the wavelengths of the Red edge-M multi-spectral sensor (Blue, Green, Red, Red edge, NIR). The values were extracted and used. The ground spectrometer measurement data for the water quality sampling point was first confirmed that the ground spectrometer measurement value was the water surface measurement value by comparing the spectral characteristics with the road material measurement data such as concrete, unpaved road, and asphalt among land cover types.

After that, correlation analysis of the terrestrial spectrometer measurement value and the UAV image value for the sampling point is performed (110) to verify that the UAV image value used in the comparative analysis of water quality and UAV image is the value of the water surface excluding the influence of the river bottom. was used in the process. The corrected multispectral images (Blue, Green, Red, Red edge, NIR) were used to produce the vegetation index NDVI, NDRE, CIRE, CVI, GNDVI, and NGRDI images used to detect vegetation and algae. The produced image is a wavelength for identifying the spectral characteristics and distribution of each water quality factor through correlation analysis with the BOD, COD, TN, TP, Chl-a, SS water quality analysis values of the samples collected after extracting the image values for each water quality sampling point. area was confirmed. Next, as a result of correlation analysis, regression analysis was performed (110) to develop a water quality prediction model for water quality items that showed a correlation in the significance level of 0.05 (p<0.05). Finally, as a result of the regression analysis, the model with the highest explanatory power among the developed water quality prediction models was selected and a water quality distribution map for each item was produced (111) through the Raster calculator tool of ArcGIS software. The prepared map was used for analysis of the possibility of understanding the distribution of water quality in small rivers, and for cause analysis (112) of the point where the concentration was high.

2 is a block diagram illustrating a water quality monitoring system using a drone according to an embodiment.

Referring to FIG. 2 , the water quality monitoring system 200 using a drone according to an embodiment may include a multi-spectral image collecting unit 210 and a water pollution level prediction unit 230 . In addition, according to an embodiment, it may further include a thermal infrared image and RGB image collecting unit 220 , a water quality factor determining unit, a green algae possibility determining unit 240 , and a water quality distribution map providing unit 250 .

First, the multi-spectral image collecting unit 210 may mount a multi-spectral sensor on the drone to acquire a multi-spectral image in an area requiring monitoring. In addition, the thermal infrared image and RGB image collecting unit 220 may additionally acquire a thermal infrared image and an RGB image through the thermal infrared camera and RGB camera of the drone.

The multispectral image collecting unit 210 produces a multispectral orthogonal image that can be used as a map of the region by matching the acquired images through image processing, and the vegetation index (NDVI, NDRE, CIRE, CVI, NGRDI, GNDVI) can be produced.

The water quality factor identification unit can identify BOD, COD, TN, TP, Chl-a, and SS factors that contribute to water pollution by performing water quality analysis by collecting water for each region. In addition, the water pollution level prediction unit 230 may analyze the correlation between multiple spectroscopy and vegetation indices and water quality factors after producing a water quality prediction model through correlation analysis (correlation analysis) and regression analysis.

As a result, in the case of the BOD item, a negative (-) correlation of -0.74 with the red wavelength was confirmed, and in the case of chlorophyll-a, NDVI, CVI, GNDVI, and NGRDI were positive (+) of 0.89, 0.76, 0.91, and 0.71, respectively. correlation can be confirmed. Finally, in the case of total nitrogen, it can be seen that NDVI and GNDVI show high positive (+) correlations of 0.94 and 0.87.

That is, since it is clear that there is a correlation between the vegetation indices, chlorophyll-a, and nutrient factors, it can be seen that chlorophyll-a and nutrient factors can be identified by inputting the vegetation indices to the model through regression analysis.

Accordingly, the green algae possibility determination unit 240 converts the multispectral image into a vegetation index, and then identifies each nutrient factor through correlation, and the water temperature distribution and RGB image obtained through a thermal infrared camera together with the nutrient factor. Through the flow of water temperature, it is possible to accurately predict the water quality status and the possibility of algal blooms before the occurrence of algal blooms.

In particular, the vegetation index can be obtained as an image, and since the thermal infrared image and the RGB image are also images, the possibility of algal blooms can be easily predicted through neural network learning. In other words, it is possible to implement a system to identify and monitor the water quality status and the possibility of algal blooms in advance through images of a wide area captured by drones.

The water quality distribution map providing unit 250 may obtain a river green algae map as an output by inputting the vegetation index image, the thermal infrared image, and the RGB image to the neural network.

3 is a flowchart illustrating a water quality monitoring method using a drone according to an embodiment.

Referring to FIG. 3 , a water quality monitoring method performed by a water quality monitoring system using a drone implemented through a computer device according to an embodiment is performed by mounting a multi-spectral sensor on the drone, thereby providing multiple monitoring for an area requiring monitoring. Acquiring a spectral image (310), and after converting the multispectral image to a vegetation index, predicting the eutrophication state and water pollution level for each point through the correlation between the multispectral image and the vegetation index and water quality factors (330) ) can be included.

Here, the method may further include the step 320 of acquiring the thermal infrared image and the RGB image through the thermal infrared camera and the RGB camera of the drone.

In addition, using the predicted eutrophication state and water pollution level at each point, the water temperature distribution identified through the thermal infrared image, and the water temperature flow identified through the RGB image, the step of identifying the water quality and the possibility of the occurrence of algae before the occurrence of algae (340) ) may be further included.

In addition, the vegetation index is obtained as an image, and inputting the vegetation index image, the thermal infrared image, and the RGB image to a neural network-based water quality prediction model may further include outputting a water quality distribution map ( 350 ).

The method may further include the step of identifying water quality factors contributing to water pollution by performing water quality analysis by collecting water for each region.

Below, each step of the water quality monitoring method using a drone according to an embodiment will be described in more detail.

The water quality monitoring method using a drone according to an embodiment may be described by taking the water quality monitoring system 200 using a drone according to the embodiment described with reference to FIG. 2 as an example.

In step 310 , the multi-spectral image collecting unit 210 may acquire a multi-spectral image for an area requiring monitoring by mounting a multi-spectral sensor on the drone. More specifically, the multi-spectral image collecting unit 210 may be produced as a multi-spectral orthogonal image that is used as a map of a region by matching the acquired multi-spectral images through image processing.

In step 320 , the thermal infrared image and RGB image collecting unit 220 may additionally acquire the thermal infrared image and the RGB image through the thermal infrared camera and RGB camera of the drone.

Meanwhile, the water quality factor identification unit may perform water quality analysis by collecting water for each region to identify water quality factors contributing to water pollution.

In step 330, the water pollution level prediction unit 230 converts the multi-spectral image to a vegetation index, and then predicts the eutrophication state and water pollution level for each point through the multi-spectral image and the correlation between the vegetation index and water quality factors. have.

More specifically, the water pollution level prediction unit 230 may analyze the multispectral image and the correlation between the vegetation index and the water quality factor through correlation analysis and regression analysis after producing a water quality prediction model. In addition, the water pollution level prediction unit 230 can identify chlorophyll-a and nutrient factors contributing to water pollution by inputting vegetation indices into the water quality prediction model produced through correlation analysis and regression analysis.

In step 340, the algae possibility determination unit 240 uses the predicted eutrophication state and water pollution level for each point, the water temperature distribution identified through the thermal infrared image, and the water temperature flow identified through the RGB image before the occurrence of algae. You can check the water quality and the possibility of algal blooms.

In step 350, the water quality distribution map providing unit 250 can output a water quality distribution map by inputting the vegetation index image, the thermal infrared image, and the RGB image to the neural network-based water quality prediction model by obtaining the vegetation index as an image. .

As described above, the embodiments can predict the eutrophication state and water pollution level for each point through the correlation between the multispectral image and the vegetation index and water quality factors acquired by drones.

In addition, in the embodiments, it is possible to determine the possibility of algae occurrence not only through the eutrophication state, but also through the water temperature status identified through the thermal infrared image and the water flow identified through the RGB image.

In addition, the embodiments may obtain a river green algae map as an output by inputting the vegetation index image, the thermal infrared image, and the RGB image to the neural network.

Hereinafter, a method and system for monitoring water quality using a drone according to an embodiment will be described.

Example

First, it is possible to collect data for monitoring water quality using a drone. Here, the data may include drone (UAV) captured images, ground spectrometer measurement, and water quality sampling data.

In this embodiment, UAV and Red edge-M multi-spectral sensors can be used, and in consideration of the characteristics of agricultural areas, various sensors such as multi-spectral sensors, LiDAR, thermal infrared, etc. This possible rotorcraft UAV can be utilized. 3DR solo (3DR Robotics), a low-cost rotorcraft UAV, was equipped with a Red edge-M (Micasense) multi-spectral sensor to shoot, and the calibration panel was photographed 2-3 times for radiation correction before flight. Next, when establishing a flight plan, after identifying the topography and features of the site through a preliminary on-site survey, an altitude of 100 m (GSD: 7.4 cm/pixcel), an overlap of 75%, a flight time of about 12 minutes (100,000 m 2/time) ) was flown.

The UAV image was acquired according to the acquisition wavelength range of the Red edge-M sensor, and consists of a total of 5 bands (Blue, Green, Red, Red edge, NIR). At one time, photos of 5 wavelengths for each band are taken, and depending on the weather conditions, about 1,500 photos per one time, a total of 18,000-21,000 photos, are carried out for about 12-14 flights including the water quality sampling point depending on the weather conditions. An image of the entire area of the target site was acquired.

PIX4D mapper software (PIX4D, Inc.) was used to map the images taken by operating the UAV. PIX4D mapper software is a program that creates high-precision maps and 3D models using 2D images such as UAV images. In addition, by using the SIFT (Scale Invariant Feature Transform) method, it is possible to maintain a constant image matching performance for photos that may be shaken due to the characteristics of the UAV used for data collection. It has the advantage of being able to correct the distortion of an image by restoring the three-dimensional coordinate data of the image, so it was used in the production of the multispectral orthogonal image of this embodiment. Multispectral orthoimage production is produced through four steps: image preprocessing, initial processing, point cloud generation, and ortho mosaic generation. In the first stage image preprocessing stage, information such as the coordinates and altitude of the photographed photo is checked, and the photo that may cause noise during orthogonal image production is removed in advance.

The two-step initial processing process is a process of creating and matching a single 3D point by automatically searching for a common point (key point) between uploaded images. In this embodiment, after the two-step initial processing process, for the geometric correction of the image, the GCP acquired in the target site in advance was input and re-optimized, and then the work was carried out step-by-step. Here, if there is a high key point between two images, the captured common area becomes larger and more common points can be matched. The conditions to become a key point are as follows. Even if the shape, size, and location of an object change, it is easy to identify, and even if the viewpoint and lighting of the sensor mounted on the UAV change, it is easy to find the corresponding point in the image. It is a corner point.

After the re-optimization process was completed, a three-step point cloud generation step was performed. Finally, an orthographic image was produced after radiation correction (Blue: 0.506, Green: 0.509, Red: 0.511, Red edge: 0.511, NIR: 0.512) based on the photos of the calibration panel taken before flight for each wavelength.

The produced multi-spectral image is a vegetation index image that can be calculated through a combination of short wavelengths as well as five wavelengths, Blue, Red, Green, Red edge, and NIR, which are basically acquired from the UAV multi-spectral sensor when identifying the water quality distribution and characteristics of small streams. was intended to be used.

As for the vegetation index, six vegetation indices such as green algae and nutrients were selected to be used for monitoring river water quality and vegetation. Selected vegetation index NDVI (Normalized Difference Vegetation Index), NDRE (Normalized Difference Red edge), CIRE (Chlorophyll Index Red edge), CVI (Chlorophyll Vegetation Index), GNDVI (Green Normalized Difference Vegetation Index), NGRDI (Normalized Green/Red) Difference Index) was produced using the Raster calculator tool of ArcGIS 10.6 software for orthoimages made in PIX4D mapper software, and the formula can be expressed as Table 1.

[Table 1]

In addition, in order to utilize multispectral images in UAV-based remote sensing techniques, it is necessary to verify the accuracy of the radiation correction performed during the production of multispectral orthographic images through terrestrial spectrometer measurement and to correct the image. Therefore, in this example, an ASD Field spec® 4 Standard-Res Spectroradiometer (Malvern Panalytical) terrestrial spectrometer was used.

The terrestrial spectrometer was measured twice on August 9 (not including the water quality sampling point) and September 30 (including the water quality sampling point), and the measurement materials were measured for rivers, vegetation, roads, yard compost, and plastic house materials. . The measurement data on land cover type was used for spectral characteristic analysis by land cover type and verification and correction of UAV multi-spectral images, and the measurement data for water quality sampling points were used to verify the appropriateness of the selection of collection points. At this time, the terrestrial spectrometer measurement at the water quality sampling point analyzes the spectral characteristics of each water sampling point and verifies that the UAV image value used in the production of the water quality prediction model excludes the influence of the river bed, thereby confirming the reliability of the developed model. (In the case of points N3, N4, and N5 where the water depth is too deep and it is difficult for an investigator wearing ground spectrometer equipment to access it, it was excluded from the measurement items). The measurement time was measured between 11:00-14:00 when the amount of insolation is high, and it was measured in consideration of weather conditions (cloudiness, etc.) and ground conditions. Next, after optimizing the sensor using RS3 software to minimize the measurement error, the White Reference was performed using a white panel with >99% reflectance for each material, followed by measurements 3-5 times.

Since the terrestrial spectrometer measurement data was acquired using RS3 software, a preprocessing process to convert the file extension (.asd) is required. In the data preprocessing process, ViewSpecPro software was used, and the .asd extension file was converted to ASCII code. After the data pre-processing, the analysis of the measured data was carried out in the following way. When verifying and correcting UAV multi-spectral images, the wavelength range of the Red edge-M multi-spectral sensor is from 350-2500 nm, which is the measurement wavelength of the terrestrial spectrometer, (Blue: 465-475 nm, Green: 550-570 nm, Red: 663-673 nm). , Red edge: 712-722 nm, NIR: 820-860 nm) and the overlapping part were extracted and the average value was calculated and used. At this time, errors determined to be caused by the measuring device or the investigator among the 3-5 measured values were excluded. Next, in order to analyze the tendency of the incident light to be reflected when the water quality sampling point and the spectral characteristics of each material were identified, the measured values were analyzed in a diagrammatic manner.

Meanwhile, as for the water quality collection point, nine points were selected considering the distribution of agricultural land, yard compost, and the confluence of rivers. This is to determine whether nutrients such as phosphorus and nitrogen actually flow into rivers due to non-point pollution factors during rainfall, and whether they cause eutrophication and algal blooms. As for the water quality sampling method, the investigator approached the center of the river in consideration of the representativeness of the river water quality, and the water was collected at a point more than about 50cm deep. It was conducted twice on August 9th and September 30th by 2,000 mL at a total of 9 points. Biochemical oxygen demand (BOD), chemical oxygen demand (COD), total nitrogen (TN), total phosphorus (TP), chlorophyll- A (Chl-a) and suspended solids (SS) were analyzed for a total of 6 items.

Next, the water quality analysis value of each point is evaluated for each item by referring to the living environment standards of rivers and living environment standards of lakes in the water quality and aquatic ecosystem environmental standards (Annexed Table of the Enforcement Decree of the Framework Act on Environmental Policy-Environmental Standards Article 2). After grading, water quality characteristics were analyzed at each point.

Next, UAV image accuracy verification and correction can be performed and a water quality prediction model can be provided.

For correlation analysis and regression analysis between the water quality factor values analyzed after water quality and the UAV image, in this embodiment, the accuracy of the UAV multispectral image was verified and corrected based on the terrestrial spectrometer measurement data, and the analysis method is as follows.

First, the value of the wavelength overlapping with the UAV multi-spectral sensor was extracted from the terrestrial spectrometer data measured for the study site, and then the UAV image value for the same point was extracted and correlation analysis was performed. In this case, as a result of the correlation analysis, the accuracy of the radiation correction performed during image production was verified by confirming the correlation between the terrestrial spectrometer measurement data and the UAV multispectral image within the significance level of 0.05. Next, in the case of multi-spectral image correction, regression models for each blue, green, red, red edge, and NIR wavelength were produced through regression analysis of UAV image values and terrestrial spectrometer data. Next, the extracted multispectral image and vegetation index image values were corrected by using the regression model.

4 is a diagram illustrating a vegetation index image obtained by converting a multi-spectral image taken by a drone according to an embodiment.

As shown in FIG. 4, when the water quality prediction model is developed, the correlation analysis of the corrected multispectral image and the vegetation index image and the water quality analysis value is performed to derive the wavelength showing the correlation for each water quality factor, so that the water quality factor is sensitive. The possibility of identifying the distribution of water quality using the response wavelength derivation and multi-spectral image was confirmed. Next, regression analysis was performed between items showing correlation within the significance level of 0.05 as a result of correlation analysis to prepare a regression model for predicting water quality distribution.

5 is a diagram illustrating a result of producing a water quality prediction model according to a regression analysis according to an embodiment.

As shown in FIG. 5 , the regression model with the highest coefficient of determination (R2) among the regression models produced as a result of the regression analysis was utilized to produce a map for water quality distribution prediction using the Raster calculator tool of ArcGIS software. At this time, only the area for the river is extracted for ease of use when the prepared map is applied to the cause analysis of the area where the concentration of water quality factors is high, and the derivation of the right-center area that should be managed first in the future water quality monitoring of small rivers. Thus, it was produced by superimposing it on the RGB image.

6 is a diagram illustrating a result of manufacturing a water quality distribution map according to an exemplary embodiment.

As shown in FIG. 6 , a water quality distribution map was prepared using the finally developed water quality prediction model, and the cause of the water quality distribution in the research target area was analyzed and the cause of the high concentration was analyzed. Also, by comparing the results of previous studies with the results of previous studies, the relevance and significance of the prediction model produced for the prediction of water quality in small rivers, where nonpoint pollution initially occurs before flowing into large and national rivers, was determined.

As described above, the water quality prediction model was produced through a total of three steps, and the results are summarized as follows. The first stage is the DB construction stage for analysis. Data were collected by operating a UAV equipped with a multi-spectral sensor, measuring the ground spectrometer, and collecting water for two periods on August 9 and September 30 at the site requiring non-point pollution control. As a result, a total of 11 images, 5 multispectral images and 6 vegetation index images, ground spectrometer measurement data for rivers, yard compost, agricultural waterways, vinyl houses, and road materials, and water quality data for 9 points were constructed. The constructed data was used to produce a water quality prediction model, which is the target in the case of UAV images and analysis values of water quality samples, and the ground spectrometer measurement data was used for verification and correction of UAV multi-spectral images and UAV image values for water quality sampling points at the bottom of the river. It was used to verify whether the influence of cotton was excluded. In particular, as a result of analyzing ground spectrometer data, in the case of agricultural waterways with low water depth, reflection characteristics similar to the road material were confirmed. It was found that the process of verifying that it is a value for sleep rather than a value is necessary.

In the second stage, the image verification and correction process, which must precede the application of the remote sensing technique, was carried out, and the terrestrial spectrometer measurement data acquired from the ground was utilized. As a result of comparative analysis of the UAV image of the terrestrial spectrometer measurement point and the terrestrial spectrometer measurement value, it was confirmed that the correlation coefficient was 0.88 (p<0.01) or higher at all wavelengths, and the regression model with an average of 80% or more of explanatory power in both periods was used. By deriving the image, verification and correction were performed. Next, as a result of analyzing ground spectrometer measurement data for water quality sampling points, it was confirmed that all water quality sampling points had reflectivity characteristics of the water body in which incident light was mostly absorbed in the near-infrared region. In addition, as a result of correlation analysis between UAV images and terrestrial spectrometer measurements for water quality sampling points, a high correlation of 0.7 (p<0.05) or more was confirmed at all wavelengths. Therefore, the UAV image value used in the production of the water quality prediction model of this embodiment is determined to be a value for water surface.

Table 2 shows the correlation coefficient between the multispectral and vegetation index water quality factors of August 9th.

[Table 2]

Table 3 shows the correlation coefficient between the multispectral and vegetation index water quality factors of September 30.

[Table 3]

In step 3, spectral characteristics analysis of water quality factors and water quality prediction model were prepared through comparative analysis of the corrected UAV image and water quality. As a result of analyzing the spectral characteristics of water quality factors, it was confirmed that Chl-a had a positive (+) correlation with the NIR wavelength, and had spectral characteristics similar to vegetation reflecting incident light in the near-infrared region. Next, TN was analyzed as having a negative (-) correlation in the visible light region (Blue, Green, Red), indicating that incident light has a high tendency to be absorbed rather than reflected. In the case of BOD, it was confirmed that as the concentration increased, incident light was mainly absorbed in the red (663-673 nm) wavelength region. By synthesizing the results of the previous analysis, a water quality prediction model was produced that can identify the distribution of BOD, TN, and Chl-a in the water system. Using the model, a map for each factor was produced to the point where water quality was not collected. Concentration distribution was identified. In the case of BOD, it was judged that it was not in the concentration range that required management because it was predicted to be above the average grade of the river living environment standard. Next, TN and Chl-a items are agricultural channels that directly flow out into the water system before and after the weir, and have high concentrations at the point where the flow rate rapidly decreases due to the curved topography, and at the point where the stream stagnates by the weir after the confluence. was predicted to have

In conclusion, according to the embodiments, it is possible to provide a water quality prediction model of three items BOD, TN, and Chl-a as a target for small rivers in agricultural areas by using UAV, ground spectrometer, and field water quality sampling data. In addition, through the examples, it was possible to identify matters to be considered when conducting research by applying the remote sensing technique to small rivers. First, the image used for remote sensing needs to be corrected through terrestrial spectrometer measurement. However, in the case of small rivers with shallow water depth, it is known that it is necessary to verify that the image value used for the analysis is the value for the water surface. could Next, when acquiring the terrestrial spectrometer and water quality sampling data of this example, there was a case in which an inaccessible area occurred due to a typhoon or torrential rain due to a narrow river width. Therefore, it is judged that it is necessary to consider in advance the terrain change caused by rainfall in addition to the depth of the river, the inflow of agricultural water, and the position of the weir when selecting the collection point. The water quality prediction model developed in this example makes it possible to identify the distribution of water quality for small streams in agricultural areas and predict the concentration at points where sampling is not carried out, so it is judged that it can be utilized as a meaningful result in the decision-making stage when managing small rivers. , and furthermore, it can be used as basic data for making water quality prediction models for long-term monitoring.

According to embodiments, when monitoring the algae phenomenon, it is possible to prevent the occurrence of algae in advance by warning of the possibility of the algal bloom before the occurrence of algae and responding according to the monitoring result. In addition, according to embodiments, it is possible to identify a nutrient factor that causes algal blooms over a wide area through drone images without on-site sampling, thereby greatly reducing the time and resources required for monitoring. In particular, it is possible to prevent algae in the target site requiring non-point pollution control. And, according to the embodiments, it is possible to accurately predict the possibility of occurrence of algae by understanding not only the nutrient factors but also the current state of water temperature and the flow of water for each area.

The device described above may be implemented as a hardware component, a software component, and/or a combination of the hardware component and the software component. For example, devices and components described in the embodiments may include, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable array (FPA), It may be implemented using one or more general purpose or special purpose computers, such as a programmable logic unit (PLU), microprocessor, or any other device capable of executing and responding to instructions. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. The processing device may also access, store, manipulate, process, and generate data in response to execution of the software. For convenience of understanding, although one processing device is sometimes described as being used, one of ordinary skill in the art will recognize that the processing device includes a plurality of processing elements and/or a plurality of types of processing elements. It can be seen that may include For example, the processing device may include a plurality of processors or one processor and one controller. Other processing configurations are also possible, such as parallel processors.

The software may comprise a computer program, code, instructions, or a combination of one or more thereof, which configures a processing device to operate as desired or is independently or collectively processed You can command the device. The software and/or data may be any kind of machine, component, physical device, virtual equipment, computer storage medium or device, to be interpreted by or to provide instructions or data to the processing device. may be embodied in The software may be distributed over networked computer systems and stored or executed in a distributed manner. Software and data may be stored in one or more computer-readable recording media.

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the medium may be specially designed and configured for the embodiment, or may be known and available to those skilled in the art of computer software. Examples of the computer-readable recording medium include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic such as floppy disks. - includes magneto-optical media, and hardware devices specially configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like.

As described above, although the embodiments have been described with reference to the limited embodiments and drawings, various modifications and variations are possible by those skilled in the art from the above description. For example, the described techniques are performed in a different order than the described method, and/or the described components of the system, structure, apparatus, circuit, etc. are combined or combined in a different form than the described method, or other components Or substituted or substituted by equivalents may achieve an appropriate result.

Therefore, other implementations, other embodiments, and equivalents to the claims are also within the scope of the following claims.

Claims (15)

In the water quality monitoring method performed by the water quality monitoring system using a drone implemented through a computer device,
Mounting a multi-spectral sensor on the drone to acquire a multi-spectral image for an area requiring monitoring; and
After converting the multi-spectral image into a vegetation index, predicting the eutrophication state and water pollution level for each point through the correlation between the multi-spectral image and the vegetation index and water quality factors
including,
acquiring a thermal infrared image and an RGB image through a thermal infrared camera and an RGB camera of the drone;
Using the predicted eutrophication state and water pollution level for each point, the water temperature distribution identified through the thermal infrared image, and the water temperature flow identified through the RGB image, determining the water quality status and the possibility of algal bloom before occurrence of algae; and
The vegetation index is obtained as an image, and outputting a water quality distribution map by inputting the vegetation index image, the thermal infrared image, and the RGB image to a neural network-based water quality prediction model.
further comprising,
After the database (DB) is built for analysis, the water quality prediction model performs the image verification and correction process that must be preceded during remote sensing, utilizes the ground spectrometer measurement data acquired from the ground, and measures the corrected drone image and water quality. By analyzing the spectral characteristics of water quality factors through comparative analysis, it is implemented through learning of the neural network,
The step of predicting the eutrophication state and water pollution level by point through the correlation is,
The multispectral image and the correlation between the vegetation index and the water quality factor are analyzed after producing a water quality prediction model through correlation analysis and regression analysis, but by inputting the vegetation index into the water quality prediction model produced through correlation analysis and regression analysis. Identification of chlorophyll-a and nutrient factors contributing to water pollution
A method for monitoring water quality, characterized in that. delete delete delete delete delete According to claim 1,
Acquiring the multi-spectral image comprises:
The acquired multi-spectral image is produced as a multi-spectral orthographic image that is used as a map of the region by matching the images through image processing.
A method for monitoring water quality, characterized in that. According to claim 1,
Identifying the water quality factors contributing to water pollution by performing water quality analysis by collecting water for each area
Further comprising a, water quality monitoring method. In the water quality monitoring system using a drone,
A multi-spectral image collecting unit that mounts a multi-spectral sensor on the drone and acquires multi-spectral images for an area requiring monitoring; and
After converting the multi-spectral image into a vegetation index, a water pollution degree prediction unit for predicting the eutrophication state and water pollution level for each point through the correlation between the multi-spectral image and the vegetation index and water quality factors
including,
a thermal infrared image and RGB image collecting unit for acquiring thermal infrared images and RGB images through the thermal infrared camera and RGB camera of the drone;
Using the predicted eutrophication state and water pollution level for each point, the water temperature distribution identified through the thermal infrared image, and the water temperature flow identified through the RGB image, the water quality status and the possibility of algal bloom before the occurrence of algal blooms. judging unit; and
The vegetation index is obtained as an image, and the vegetation index image, the thermal infrared image, and the RGB image are input to a neural network-based water quality prediction model to output a water quality distribution map.
further comprising,
After the database (DB) is built for analysis, the water quality prediction model performs the image verification and correction process that must be preceded during remote sensing, utilizes the ground spectrometer measurement data acquired from the ground, and measures the corrected drone image and water quality. By analyzing the spectral characteristics of water quality factors through comparative analysis, it is implemented through learning of the neural network,
The water pollution degree prediction unit,
The multispectral image and the correlation between the vegetation index and the water quality factor are analyzed after producing a water quality prediction model through correlation analysis and regression analysis, but by inputting the vegetation index into the water quality prediction model produced through correlation analysis and regression analysis. Identification of chlorophyll-a and nutrient factors contributing to water pollution
Characterized in, the water quality monitoring system. delete delete delete delete delete 10. The method of claim 9,
The multi-spectral image collecting unit,
The acquired multi-spectral image is produced as a multi-spectral orthographic image that is used as a map of the region by matching the images through image processing.
Characterized in, the water quality monitoring system.

Images