KR101933216B1 - River topography information generation method using drone and geospatial information - Google Patents

Abstract

The present invention relates to a method of generating a river terrain information using a drone and spatial information, and more particularly, to a method of generating a DEM by processing gradient information and an area expansion method using spatial information technology from 3D DSM data acquired by a drone DEM images were obtained through image matching and orthogonal images and DSM data were constructed. Since the constructed DSM data includes the vegetation and trees in the stream, Based on the point cloud data acquired during the UAV photographing, we applied the gradient method and selected 100 verification points. Then, using the VRS (Virtual Reference Station) and the total station survey, The reliability of DEM data by the drone point cloud-based gradient method by evaluating the accuracy of the data It provides.

Description

[0001] The present invention relates to a method of generating a river topography information using a drone and spatial information,

The present invention relates to a method for generating a stream topographical information using a drone and spatial information. More particularly, the present invention relates to a method for generating a stream topographical information by using a gradient information technique and a region expansion technique from a 3D DSM (Digital Surface Model) The present invention relates to a method of generating a stream topographical information using a drone and spatial information, which develops a technique of generating a DEM (Digital Elevation Model)

1. BACKGROUND ART

The topographic data for the surrounding area of rivers are essential for the establishment of comprehensive river basins such as national rivers, regional rivers and small rivers, as well as limiting the environment of rivers, irregular development activities, and disaster management such as floods. Recently, various studies using spatial information and image data have been attempted in river works such as basic river maintenance plan, river facility installation, flood analysis, and analysis of similarity.

First, there is a study on the effect of inundation due to the density of LiDAR data using GIS, including the study of the uniform cross section and the finite element network by using actual cross section in GIS - based software. In addition, there have been studies to map 3D river boundaries using spatial information, and to extract the shape of stream through image classification method from satellite images.

In order to obtain the topographic data of rivers, we conducted a survey on river conditions, regional area, and topography using 1 / 50,000 topographic map. Then, GPS (Global Positioning System), total station, Level and longitudinal cross-sectional surveying. Since the conventional river survey method has a limited time and cost to observe a wide range of river areas, the horizontal position and elevation value are observed at about 200 transversal intervals and the output is calculated. Therefore, it is not possible to model the shape of continuously changing stream in three dimensions. Especially, when the irregular stream is distributed much in topographic features such as vegetation and trees, it is difficult to reflect the exact elevation value of the ground. In this way, since the river terrain data depending on the cross-sectional survey results can not represent the topography of the entire river, there is a problem that it is difficult to secure the reliability in the flood modeling and the similarity evaluation as well as the determination of the scale facility size through the river maintenance basic plan . Therefore, it is required to study the rapid and reliable 3D terrain data generation for rivers.

In recent years, research on the construction of 3D terrain data using drones has been actively conducted in the field of spatial information. The drones are called unmanned aerial vehicles (UAVs). The drones are equipped with GPS and INS (Inertial Navigation System) sensors to acquire position information and attitude information, as well as a camera that captures images. In addition, the drones are equipped with a VRS (Virtual Reference Station) So that we can model the terrain faster.

A study using drones for the modeling of topographical landforms around rivers is as follows. First, there is a study that evaluated the utility of low - altitude terrain data using drones (UAV), and also analyzed the DSM (Digital Surface Model) and accuracy of waterfront structures from drones. In addition, there have also been studies on modeling of beach topography using drones or using drone photogrammetry to acquire topographical data for inaccessible areas. As a practical example of the disaster disaster area, a study was conducted to evaluate the accuracy of water detection by applying various image classification techniques to images acquired after mounting a near infrared ray (NiR) sensor on a drone, There have also been studies evaluating landslide risk in connection with analytical models.

As a result, many researches have been made on 3D terrain modeling using drones and disaster disaster, urban planning, vegetation, and agriculture. However, in the case of terrain modeling using existing drones, it is based on DSM data including all of the features. Since the topography data used in the flood modeling and sediment analysis, including the basic plan for river maintenance, is the digital Elevation Model (DEM) data, which is the elevation value of the pure ground, the river topography has various features such as vegetation, trees, DEMs can be obtained by removing features from DSM data.

 C., Bebe, R. W., McLain, T. W. and Mehra, R. K., 2008, "Fire Monitoring with Multiple Small UAVs", Portland, USA. Colomina.  Colomina, I., Blazquez, M., Molina, P., Pares, M.E. and Wis, M., 2008, "A New Paradigm for High-Resolution Low-Cost Photogrammetry and Remote Sensing" ISPRS XXI Congress. Beijing, China, Jul. 3-11, 2008. pp.1201-1206.  Everaerts. J., 2008, "Use of Unmanned Aerial Vehicles (UAVS) for Remote Sensing and Mapping" ISPRS XXI Congress, 2008.6. Beijing China, pp. 1187-1191.  Grenzd Orffer, G. J., Engel, A., Teichert, B., 2008, "Photogrammetric Potential of Low-Cost UAVS in Forestry and Agriculture" ISPRS XXI Congress, 2008.6. Beijing China, pp. 1207-1213.  Nagai, M., Chen, T., Ahmed, A. and Shibasaki, R., 2008, "Borne Mapping by Multi Sensor Integration" ISPRS XXI Congress, Beijing, China, Jul. 3-11, 2008, pp.1215-1221.  "Light-Weight Multispectral Sensor for Micro UAV - Opportunities for Very High Resolution Airborne Remote Sensing" ISPRS XXI Congress, Beijing, China, Jul., 2008. Nebiker, S., Annen, A., Scherrer, M. and Oesch, . 3-11, 2008. pp.1193-1199.

It is an object of the present invention to solve the problems of the prior art, and it is an object of the present invention to develop a technique of generating a DEM by processing gradient information and area expansion method using spatial information technology from 3D DSM data acquired by a drone, And the DSM data were constructed through the image matching. The DSM data constructed in this study includes the vegetation and the trees in the stream. Therefore, in order to generate the DEM based on the stream ground, Based on the point cloud data based on the gradient method, 100 points of verification were selected, and then the measurement values calculated from the VRS (Virtual Reference Station) survey and the total station survey were used to calculate the elevation accuracy of the DSM and DEM data , Which provides reliability of DEM data by the drone point cloud-based gradient method And using spatial information to provide a stream terrain information generation method.

In order to achieve the object of the present invention, a method for generating a stream topographical information using a drone and spatial information comprises the steps of: (a) providing a virtual reference station (VRS) to a GCP (Ground Control Point) A reference point) survey; (b) acquire k camera images of drone (UAV), GPS and INS information are connected to each photo file, and matching sheets by image matching software (Pix4D SW) Generating a drone-based orthoimage and a DSM (Digital Surface Model) by improving the accuracy of the 3D point cloud from the INS information; (c) Since the digital surface model (DSM) includes vegetation and trees in the stream, the digital elevation model (DSM) of the dron (UAV) is used to generate a digital elevation model (DEM) A step of generating a DEM reflecting the elevation values available for the river surveying using the gradient method and the area expansion method based on the point cloud data from the 3D DSM data obtained through the photographing; (D) Apply gradient method based on point cloud data obtained from drones (UAV) camera shooting, select m verification points, and then use measurement values calculated from VRS and total station survey to calculate DSM And evaluating the horizontal and vertical accuracy of the DSM data by the drone point cloud-based gradient technique by evaluating the elevation accuracy of the DEM data.

The drones are fixed-wing drones equipped with NiR cameras in airplane shape. The drones are designed to shoot at a flight height of 195 m at a flight altitude of 6 ㎝, and the longitudinal and lateral redundancy are designed to be 85% and 70%, respectively. .

In the step (a), the image obtained through camera shooting of the drone is basically a WGS84 UTM coordinate system. In order to convert it into GRS80 TM, which is a domestic geodetic coordinate system, 10 GCPs (Ground Control Point, And a VRS (Virtual Reference Station) survey is performed on 10 GCPs.

Based on the point cloud data obtained through the camera shooting of the drones, a grading method is used to generate a DEM reflecting the elevation values available for stream surveying. In order to classify the trees in the stream using point cloud data, In order to extract the ground points, we first investigate the slope of the target area, calculate the slope using the distance between the point cloud data and the elevation difference based on the slope, and then calculate the slope of the ground And dividing the ground points into a plurality of point clouds, dividing cells having a predetermined size, dividing the points within a given inclination into ground points, and calculating an elevation among a plurality of point clouds existing in the cell after the virtual cells are generated We assign it to the lowest point seed pixel and calculate the slope with the neighboring cell The user applies a region growing method extends out of the cell within the threshold slope value specified.

The region expansion method using point cloud data is a theory developed in the image field. Generally, pixels belonging to the same object region in an image are based on having similar statistical characteristics. Using these characteristics, similar regions are expanded by comparing them with surrounding pixel values starting from an initial starting point, i.e., a seed pixel, and comparing the intensity values of neighboring pixels with the seed pixel, And,

(1)The process of examining the homogeneity of regions in an image is examined by Equation (1)

(One)

는 초기 seed 픽셀을 의미하고,

는 초기 seed 픽셀과 이웃한 k개의 픽셀들을 의미하고, N은 이웃한 픽셀들의 전체 개수이며, 만약에 초기 seed 픽셀과 이웃한 영역이 서로 동질성을 갖는다면 식 (1)을 만족하게 되어 한 영역으로 확장되며, Here, P (-) is called a logical predicate, and if True, the area in parentheses becomes the same area,

Denotes an initial seed pixel,

Is the number of neighboring pixels, and N is the total number of neighboring pixels. If the initial seed pixel and neighboring regions are homogeneous, then Equation (1) is satisfied. Expanded,

Based on the Microstation V8 and the Terra Scan / Terra Modeler terrain data processing software, a digital elevation model (DEM) generation process based on the slope technique was applied. The parameters applied in the preprocessing process are density of the point cloud, It depends on the terrain condition. Some values are adjusted based on the parameter values commonly used in the existing studies, and the DEM is selected by visual inspection.

In the preprocessing step, all points are classified into a default class, and points having an elevation value that is too low or too high as compared with other points are designated as a low point class. Low point class classification is performed and the remaining default class The ground class is basically set by setting a maximum angle that can be generated, and each point is navigated through an inclination. In order to extract the ground class, a radius of 2 m The maximum distance that can be searched once was limited to 5m, and the process of finding the other points repeatedly based on the angle within the limited area was performed, and classification of ground points was performed. All points in the Default Class are classified as Low Vegetation Class and Low Vegetation Class When there is more than 0.5m books that are based on the fact that they exist as a Medium vegetation Class, more than 3m, and are classified as High vegetation Class,

In order to classify the misclassified points into the ground points by checking again from the orthoimage image, basically, since the ground is classified on the basis of 6 °, Is used to modify the land surface by using Terra Scan topographical data processing software to prevent erroneous errors recognized as trees in the vicinity of the rivers.

In the above method, the elevation distribution of DSM and DSM in the transverse section is analyzed for the elevation values of DSM and DEM for the transverse section by selecting a part of the stream including vegetation and trees. The elevation distribution of the DSM is higher than that of the DEM generated by using the slope method in which the elevation value of the DSM is located. In contrast, in the case of the DEM data generated by the gradient method based on the point cloud, Is similar to the elevation value of the ground.

In this method, 100 verification points were selected for the evaluation of the accuracy of the DEM data generated through the DSM data obtained through the drone shooting with the camera and the point cloud based gradient method. The verification points include the vegetation-free topography Choose from a wide range of heights, ranging from ground to high vegetation or trees.

In the step (c), the elevation of each verification point is obtained by performing a VRS (Virtual Reference Station) survey and a total station survey, and the horizontal position and elevation value are obtained through the VRS survey, In the case of a tree with a large height from the ground, it is difficult to perform the VRS survey. Therefore, the total station is used to locate the prism on the adjacent terrain and then obtain the elevation value by the remote altitude surveying method.

Each elevation elevation for each verification point was obtained by performing VRS surveying and total station surveying. Horizontal position and elevation value were obtained through VRS (Virtual Reference Station) surveying of the terrain with little vegetation. , It is difficult to perform the VRS survey. Therefore, the total station survey was used to locate the prism on the adjacent terrain,

DSM, DEM, and maximum (TRUE) elevation distribution for 100 verification points. As a result of the analysis, the DEM generated by the gradient method from the point cloud data showed a pattern similar to the maximum value (TRUE), whereas the DSM showed similar to the maximum value (TRUE) at the verification point with little vegetation .

The elevation error is calculated based on the maximum value (TRUE) obtained from the VRS survey and total station survey for DSM and DEM data by 100 verification points, and the height error range of the DSM based on the maximum value (TRUE) 0.08 ~ 15.33m, and the standard error was calculated as ± 3.84m. On the other hand, the height error range of DEM was 0.27 ~ 0.41m and the standard error was calculated as ± 0.12m. It was confirmed that the error was low,

In the case of stream survey where the elevation value of the actual ground should be reflected, the point cloud based gradient method is applied to the DEM from the DSM data obtained from the camera image of the drone (UAV)

The DEM data analyzed by applying the UAV point cloud-based gradient method are used as the topographic data for the river maintenance basic plan, flood modeling, and similarity assessment.

According to the present invention, a method of generating a DEM by processing a gradient information and an area expansion method using a spatial information technology from a 3D DSM data acquired by a drone was developed. The drones were acquired and the orthoimages and DSM data were constructed through image matching. Since the constructed DSM data includes the vegetation and the trees in the stream, the slope method is applied based on the point cloud data acquired during the UAV photographing to generate the DEM based on the stream ground. Then, after selecting 100 verification points, the accuracy of the DSM and DEM data is evaluated by using the measurement values obtained from the VRS (Virtual Reference Station) survey and the total station survey, so that the drone point cloud-based slope method The reliability of the DEM data is presented.

In this study, the geomorphological data that can be used for the river service were constructed by applying the gradient method from the point cloud data acquired through the drones survey. The results of the study were as follows; .

First, the flight plan was established using eMotion SW flight planning software, and 391 photographs obtained from the drone shooting were processed with Pix4D SW image matching software, and the technique of constructing orthophoto, point cloud and DSM data could be presented . Especially, the standard deviations of X (E), Y (N), and Z were 0.016m, 0.019m, and 0.041m, respectively, , Point cloud, and DSM data within 2 ㎝ and 5 ㎝, respectively.

Secondly, based on the point cloud obtained from UAV, it was possible to construct the DEM data that can be applied to the river work by applying the slope technique. Especially, It was confirmed that the tree height was removed from DEM.

Third, we obtained the maximum value (TRUE) for 100 verification points by using VRS survey and total station survey. As a result, the standard errors of DSM and DEM were calculated to be ± 3.84m And ± 0.12m, respectively. Therefore, it was found that DEM data analyzed by UAV point cloud - based gradient method is effective as topographic data for stream work such as river maintenance basic plan, flood modeling, and sediment valuation.

FIG. 1 is a photograph of a study site for generating a drone-based orthoimage and DSM.
FIG. 2 is a screen for performing a VRS (Virtual Reference Station) survey on GCPs 2, 9, and 10 (ground control points).
FIG. 3 is a screen for designing a drones flight plan using the flight planning software (eMotion SW).
4 is a screen for connecting GPS and INS information to a photo file using the " Flight Data Manager " function of the eMotion SW.
FIG. 5 is a screen showing image fusion processing by using image matching software (Pix4D SW) of Drones aerial photography photographs.
6 is a photograph showing Orthomosaic and DSM (Digital Surface Model) data generated through the image matching software (Pix4D SW).
FIG. 7 is a three-dimensional modeling screen for some river sections implemented by Virtual Survey SW using an orthophoto image and a DSM file.
8 is a delay point classification diagram through cell division.
9 is a schematic diagram of the area expansion method.
10A is a flowchart showing a digital elevation model (DEM) generation process based on the gradient technique.
Fig. 10B is a view for explaining a point cloud (point cloud data).
10C is a view for explaining the total station.
Fig. 11 is (a) Dron DSM data cut into boundaries set for the accuracy evaluation of river terrain data, and (b) DEM generated using point cloud data based gradient method.
FIG. 12 is a view showing the elevation values of DSM and DEM for a transverse section by selecting a part of the area of Hicheon including vegetation and trees.
13 is a verification point survey photograph utilizing a VRS (Virtual Reference Station) and a total station.
FIG. 14 is a graph showing DSM, DEM, and maximum value (TRUE) elevation distribution per 100 verification points.
15 is a graph showing DSM and DEM errors for each verification point.
16 is a diagram comparing DSM and DEM standard error analysis.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

2. Drone-based orthoimage and DSM generation

2.1 Site selection and ground reference point survey

The purpose of this study is to develop a method to generate DEM by processing gradient information and area expansion method with spatial information technology from 3 - D DSM data acquired by drone. In this study, some sections of Gapcheon located in Seo - gu of Daejeon Metropolitan City were selected as study sites and drones were acquired. Orthographic images and DSM data were constructed through image matching. Since the constructed DSM data includes the vegetation and the trees in the stream, the slope method is applied based on the point cloud data acquired during the UAV photographing to generate the DEM based on the stream ground. Then, after selecting 100 verification points, the accuracy of DSM and DEM data is evaluated by using the measurement values obtained from the VRS (Virtual Reference Station) survey and the total station survey, so that the drone point cloud- The reliability of the DEM data was suggested.

FIG. 1 is a photograph of a study site for generating a drone-based orthoimage and DSM.

In the embodiment, as shown in FIG. 1, some river sections located in the western part of Daejeon Metropolitan City are selected as study sites. The target area is the Gap River section of the national river, which has a length of about 1.5 km. The river terrain includes various types of terrain such as vegetation, trees, and structures.

The images obtained through drone shooting will have the WGS84 UTM coordinate system basically, and the ground control point (GCP) survey must be performed for conversion to the domestic geodetic coordinate system GRS80 TM. As shown in FIG. 1, a total of 10 points were selected as the ground reference point (GCP) for the GCP survey, and the GCP survey was used for the VRS survey capable of positioning using the virtual reference network. The VRS survey has a horizontal and vertical error of about 3 ~ 5 cm. However, the river terrain survey is very irregular and the VRS survey is used because it causes a considerable error depending on the location of the receiver. Table 1 shows the results of the GCP survey obtained through the VRS survey, and FIG. 2 shows the VRS (Virtual Reference Station) survey for the GCPs 2, 9, and 10 (ground control points) .

2.2 Drone-based Orthographic Image and DSM Generation

The drone equipment used for river terrain data acquisition is an airplane-shaped fixed wing dron (eBee model manufactured by SenseFly, Switzerland) that can fly up to 50 minutes and can fly about 12 km2 on a single flight. The digital camera is a WX220 model of Sony, a non-surveying camera, with a focal length of 4 to 25 mm.

Table 2 shows the specifications of the drones used in this study.

The flight plan of the drone (UAV) was made using eMotion SW and the target area included high mountain area nearby, so it was designed to shoot at a flight level of 195m and resolution of 6cm in consideration of the risk of collision when shooting the drones Respectively. In addition, the longitudinal and lateral redundancy was designed to be 85% and 70%, respectively.

FIG. 3 is a screen for designing a drones flight plan using the flight planning software (eMotion SW).

A total of 391 photographs were photographed for the study area and the GPS and INS information were connected to the photograph file as shown in FIG. 4 by using the "Flight Data Manager" function of the eMotion SW.

FIG. 5 is a screen showing image fusion processing by using image matching software (Pix4D SW) of Drones aerial photography photographs.

In the embodiment, as shown in Fig. 5, image registration was performed using Pix4D SW (image registration software). The standard deviations of X (E), Y (N), and Z for 10 GCP (ground reference points) obtained by VRS survey were 0.016m, 0.019m, and 0.041m, respectively. Therefore, the horizontal and vertical accuracy of orthoimage, point cloud, and DSM data can be secured within 2 ㎝ and 5 ㎝, respectively.

The image matching software (Pix4D SW) improves 3D point cloud accuracy from photograph, GPS, INS (Inertial Navigation System) information, and ultimately produces highly reliable orthographic image and DSM data.

6 is a photograph showing Orthomosaic and DSM (Digital Surface Model) data generated through the image matching software (Pix4D SW). For reference, boundaries and validation points for the accuracy evaluation of river terrain data are shown in the orthoimage image. The capacity of the orthoimage and DSM data are 837MB and 816MB, respectively.

3. Construction of river terrain data using spatial information technology

Digital Surface Model (DSM) data constructed by using a camera of a drone (UAV) includes vegetation and trees in the river. Therefore, the DEM (Digital Elevation Model, ) Data.

FIG. 7 is a three-dimensional modeling screen for some river sections implemented by Virtual Survey SW using an orthophoto image and a DSM file. It is shown that the DSM data reflecting elevation values higher than the actual ground were generated due to the influence of trees and vegetation in the stream.

Studies using point cloud data to extract or remove these features have been centered around aerial laser surveying. Previous studies using point clouds have focused on extracting buildings using the Local Maxima method. The Local Maxima method uses the size and height threshold of the filter as a variable and calculates the lowest elevation value of the point data contained in the filter and classifies the data having elevation value higher than the assumed elevation value as the non-elevation point . Local Maxima needs to know in advance the maximum building size of the ground such as the building to be classified in the applicable area. In order to remove very large ground water, it becomes difficult to apply in the slope area because the filter becomes larger in size and at the same time the slope of the filter becomes smaller.

Therefore, in the present invention, a DEM reflecting the elevation values available for stream surveying is generated using the gradient method based on the point cloud data acquired through the camera shooting of the drones.

In order to classify the trees in the stream using the point cloud data, the process of distinguishing between the non-ground point and the ground point should be preceded. In order to extract the land surface, we first investigate the slope of the target area, calculate the slope using the distance between the points in the point cloud data and the elevation difference based on the slope, compare with the slope of the ground, and classify the ground point shall. In this case, if the area of the calculated cell is too large, the problem of misclassifying the tree as shown in Fig. 8 (a) occurs. To solve this problem, the area of the cell considering the width and topographic characteristics of the trees It is important to set it up.

8 is a delay point classification diagram through cell division. 8 (b) is a process of dividing a cell having a predetermined size into points corresponding to a given inclination, and FIG. 8 (a) shows that the problem of misclassifying the trees into the ground is solved.

After a virtual cell is created, it is assigned to the lowest point cloud among the many point clouds existing in the cell, and the cell is expanded within the threshold slope specified by the user by calculating the slope with the neighboring cell Outgoing area extension method was applied.

9 is a schematic diagram of the area expansion method.

The area expansion method using point cloud data is a theory developed in the image field. In general, the pixels belonging to the same object region in the image have a similar statistical characteristic. The region expansion method is a method of extending the similar region by comparing these values with surrounding pixel values starting from an initial starting point, i.e., a seed pixel, using this characteristic of the image. As shown in FIG. 8, the seed pixel is compared with the intensity values of neighboring pixels, and when the similarity is high, the report area is expanded to the same area (Pitas, 1993).

The process of examining the homogeneity of a region in an image is shown in Equation (1).

(1)

(One)

는 초기 seed 픽셀을 의미하고,

는 초기 seed 픽셀과 이웃한 k개의 픽셀들을 의미한다. 여기서 N은 이웃한 픽셀들의 전체 개수이다. 만약 초기 seed 픽셀과 이웃한 영역이 서로 동질성을 갖는다면 식 (1)을 만족하게 되어 한 영역으로 확장될 수 있다.Here, P () is called a logical predicate. If True, the area in parentheses is the same area.

Denotes an initial seed pixel,

Means the initial seed pixel and neighboring k pixels. Where N is the total number of neighboring pixels. If the initial seed pixel and neighboring regions are homogeneous, the region can be extended to one region by satisfying Equation (1).

In the present invention, a digital elevation model (DEM) generation process based on the gradient technique as shown in FIG. 10 is applied based on Microstation V8 of Bentley and Terra Scan / Terra Modeler of Terrasolid. The parameters applied in the preprocessing process may vary depending on the density of the point cloud and the terrain conditions of the site. Therefore, in the present invention, some values were adjusted based on the parameter values generally used in the existing studies, and the most appropriate DEM was selected through visual inspection.

In the preprocessing process, all the points are classified into the Default Class, and points having an elevation value that is too low or too high compared to other points are designated as the Low Point Class. The Low Point Classification is performed and the ground class is classified for the remaining Default Classes. The ground class is basically performed by setting the maximum angle that can be generated, and navigating each point through the gradient. In the present invention, other points existing within a radius of 2 m are searched based on a commonly used angle of 6 ° for extracting a ground class. The maximum distance that can be retrieved once was limited to 5m, and the process of finding different points repeatedly based on the angle within the limited area was performed. Classification of ground points was performed and all remaining points of the Default Class were classified as Low vegetation class. In the low vegetation class, the points that exist more than 0.5m from the ground point are classified as Medium vegetation class and the points existing more than 3m are classified as High vegetation class.

The ground points to be utilized in the present invention are checked again from the orthoimage image, and the misclassified points are classified into ground points. Basically, since the ground is classified on the basis of 6 °, the vicinity of the riverside bank bank where the steep slopes are appeared leads to errors recognized as trees. Therefore, in the present invention, the ground point is corrected by setting Terra Scan terrain data processing software as a separate area near the river bank.

In the present invention, the boundary is set for the accuracy evaluation of the stream topographical data. The boundaries were determined based on river bank in the case of the left eye tributary and in the case of the right eye untreated area in consideration of the extent of the flood.

10A is a flowchart showing a digital elevation model (DEM) generation process based on the gradient technique.

The method of generating a river terrain information using the drones and spatial information according to the present invention comprises the steps of: (a) performing a VRS (Virtual Reference Station) survey on n GCPs (ground control points, ground reference points) step; (b) acquire k camera images of drone (UAV), GPS and INS information are connected to each photo file, and matching sheets by image matching software (Pix4D SW) Generating a drone-based orthoimage and a DSM (Digital Surface Model) by improving the accuracy of the 3D point cloud from the INS information; (c) Since the digital surface model (DSM) includes vegetation and trees in the stream, the digital elevation model (DSM) of the dron (UAV) is used to generate a digital elevation model (DEM) A step of generating a DEM reflecting the elevation values available for the river surveying using the gradient method and the area expansion method based on the point cloud data from the 3D DSM data obtained through the photographing; (D) Apply gradient method based on point cloud data obtained from drones (UAV) camera shooting, select m verification points, and then use measurement values calculated from VRS and total station survey to calculate DSM And evaluating the horizontal and vertical accuracy of the DSM data by the drone point cloud-based gradient technique by evaluating the elevation accuracy of the DEM data.

The drones are fixed-wing drones equipped with NiR cameras in airplane shape. The drones are designed to shoot at a flight height of 195 m at a flight altitude of 6 ㎝, and the longitudinal and lateral redundancy are designed to be 85% and 70%, respectively. .

In the step (a), the image obtained through camera shooting of the drone is basically a WGS84 UTM coordinate system. In order to convert it into GRS80 TM, which is a domestic geodetic coordinate system, 10 GCPs (Ground Control Point, And a VRS (Virtual Reference Station) survey is performed on 10 GCPs.

Based on the point cloud data obtained through the camera shooting of the drones, a grading method is used to generate a DEM reflecting the elevation values available for stream surveying. In order to classify the trees in the stream using point cloud data, In order to extract the ground points, we first investigate the slope of the target area, calculate the slope using the distance between the point cloud data and the elevation difference based on the slope, and then calculate the slope of the ground And dividing the ground points into a plurality of point clouds, dividing cells having a predetermined size, dividing the points within a given inclination into ground points, and calculating an elevation among a plurality of point clouds existing in the cell after the virtual cells are generated We assign it to the lowest point seed pixel and calculate the slope with the neighboring cell The user applies a region growing method extends out of the cell within the threshold slope value specified.

The region expansion method using point cloud data is a theory developed in the image field. Generally, pixels belonging to the same object region in an image are based on having similar statistical characteristics. Using these characteristics, similar regions are expanded by comparing them with surrounding pixel values starting from an initial starting point, i.e., a seed pixel, and comparing intensity values of neighboring pixels with the seed pixel, if the similarity is high (high degree of similarity) , Enlarging the area,

(1)The process of examining the homogeneity of regions in an image is examined by Equation (1)

(One)

는 초기 seed 픽셀을 의미하고,

는 초기 seed 픽셀과 이웃한 k개의 픽셀들을 의미하고, N은 이웃한 픽셀들의 전체 개수이며, 만약에 초기 seed 픽셀과 이웃한 영역이 서로 동질성을 갖는다면 식 (1)을 만족하게 되어 한 영역으로 확장되며, Here, P (-) is called a logical predicate, and if True, the area in parentheses becomes the same area,

Denotes an initial seed pixel,

Is the number of neighboring pixels, and N is the total number of neighboring pixels. If the initial seed pixel and neighboring regions are homogeneous, then Equation (1) is satisfied. Expanded,

Based on the Microstation V8 and the Terra Scan / Terra Modeler terrain data processing program, a digital elevation model (DEM) generation process based on the gradient technique was applied. The parameters applied in the preprocessing process are the density of the point cloud, It can be varied depending on the geographical conditions, and some values are adjusted based on the parameter values commonly used in the existing studies, and the most suitable DEM is selected through visual inspection.

Fig. 10B is a view for explaining a point cloud (point cloud data).

The point cloud (point cloud data) can extract a point cloud having three-dimensional coordinates (X, Y, Z) using photogrammetry. As a type of 3D photogrammetry calculation method, a height value of a feature point is calculated from two photographs taken in duplicate for the same point in a manner similar to triangulation.

10C is a view for explaining the total station.

The total station is equipped with the ability to simultaneously observe the angle and the distance. It is a device that can calculate the coordinates of a new point to be measured when two coordinates (X, Y, Z) are known. It is an electronic measuring instrument that can calculate and display the data in a short time in the microprocessor installed in the total station, and convert the result to a computer.

Fig. 11 is (a) Dron DSM data cut into boundaries set for the accuracy evaluation of river terrain data, and (b) DEM generated using point cloud data based gradient method. In addition, the validation for verifying the accuracy of the stream topography data is shown in Figs. 11 (a) and (b).

FIG. 12 is a view showing the elevation values of DSM and DEM for a transverse section by selecting a part of the river including vegetation and trees as a target area. As a result of the analysis of the elevation distribution of DSM data in the transverse section, it can be seen that the elevation value of DSM is much higher than that of DEM which shows the ground because the trees with a height of 10m or more are located. On the other hand, the DEM data generated by the gradient method based on the point cloud showed an elevation distribution of about 63 ~ 64 EL.m, and thus it was confirmed that it is similar to the elevation value of the ground.

In order to evaluate the accuracy of the DSM data obtained through the drone shooting with the camera and the DEM data generated by the point cloud based gradient method, 100 verification points were selected as shown in FIGS. 11 (a) and (b). The verification points were selected to include various terrains with little vegetation, as well as vegetation and trees with a large height from the ground.

13 is a verification point survey photograph utilizing a VRS (Virtual Reference Station) and a total station.

As shown in FIG. 13, each verification point elevation was obtained by performing a VRS surveying and a total station surveying. The VRS (Virtual Reference Station) survey was used to obtain the horizontal position and altitude of the terrain with little vegetation. In the case of trees with a large height from the ground, it is difficult to perform the VRS survey. After the prism was placed on the terrain, elevation values were obtained by the remote altimeter survey method.

FIG. 14 is a graph showing DSM, DEM, and maximum value (TRUE) elevation distribution per 100 verification points. As a result of the analysis, the DEM generated from the point cloud data using the gradient method showed a pattern similar to the maximum value (TRUE). On the other hand, DSM showed a similarity to the maximum value (TRUE) at the verification point with little vegetation, but a large difference from the maximum value (TRUE) at the verification point with high plant height.

15 is a graph showing DSM and DEM errors for each verification point.

The elevation error was calculated based on the maximum value (TRUE) obtained from the VRS survey and the total station survey for DSM and DEM data by 100 verification points as shown in FIG. The height error range of the DSM based on the maximum value (TRUE) was 0.08 ~ 15.33m and the standard error was calculated as ± 3.84m. On the other hand, the height error range of DEM is 0.27 ~ 0.41m, and the standard error is ± 0.12m, which shows that the altitude error is much lower than DSM data. Therefore, in the case of stream survey which needs to reflect the actual ground level, it is effective to generate and utilize DEM by applying point cloud based gradient method from DSM data obtained from camera image of drone (UAV).

16 is a diagram comparing DSM and DEM standard error analysis.

As a result, the present invention constructed the terrain data that can be utilized for the river business by applying the gradient method from the point cloud data obtained through the survey of the drones, and conducted a research by selecting a section of the Gapcheon area in Daejeon, The results were as follows.

First, flight plan was established using dron flight planning software, eMotion SW, and 391 photographs obtained from the camera of the dron (UAV) were processed by Pix4D SW (image matching software), and ortho image, point cloud and DSM data And the method of constructing the. Especially, the standard deviations of X (E), Y (N), and Z were 0.016m, 0.019m, and 0.041m, respectively, as a result of GCP survey to ensure the reliability of video connection of 391 drone camera photographs. Therefore, the horizontal and vertical accuracy of the final result, orthoimage, point cloud, and DSM data, could be secured within 2 ㎝ and 5 ㎝, respectively.

Secondly, based on the point clouds acquired through drone (UAV), it was possible to construct the DEM data that can be applied to the river work by applying the slope technique. Especially, the cross - The tree height was removed from the DEM.

Third, we obtained the maximum value (TRUE) for 100 verification points by using VRS surveying and total station. As a result, the standard errors of DSM and DEM were calculated to be ± 3.84m ± 0.12m. Therefore, it was found that DEM data analyzed by UAV point cloud - based gradient method is effective as topographic data for stream work such as river maintenance basic plan, flood modeling, and sediment valuation.

As described above, the method of the present invention can be implemented as a program and recorded on a recording medium (CD-ROM, RAM, ROM, memory card, hard disk, magneto-optical disk, storage device, etc.) Lt; / RTI >

While the present invention has been described in connection with what is presently considered to be practical exemplary embodiments, it is to be understood that the invention is not limited to the disclosed embodiments, but, on the contrary, It will be understood that various modifications and changes may be made without departing from the scope of the present invention.

1-10: GCP: Ground Control Point, Ground Control Point
VRS: Virtual Reference Station, Virtual Reference Point
DSM: Digital Surface Model, Numerical Surface Model
DEM: Digital Elevation Model, Digital Elevation Model

Claims (11)

(a) performing a VRS (Virtual Reference Station) survey on n GCPs (ground control points) at a destination of the stream terrain;
(b) acquire k camera images of drone (UAV), GPS and INS information are connected to each photo file, and matching sheets by image matching software (Pix4D SW) Generating a drone-based orthoimage and a DSM (Digital Surface Model) by improving the accuracy of the 3D point cloud from the INS information;
(c) The digital surface model (DSM) includes the vegetation in the river and the trees, so it is necessary to use the digital elevation model (UEM) A step of generating a DEM reflecting the elevation values available for the river surveying using the gradient method and the area expansion method based on the point cloud data from the 3D DSM data obtained through the photographing; And
(d) Apply a slope method based on the point cloud data acquired during the drone (UAV) camera shooting, and select m verification points. Then, using the measurement values calculated from the VRS survey and the total station survey, Evaluating the horizontal and vertical accuracy of the DSM data by the drone point cloud-based gradient technique by evaluating the elevation accuracy of the data;
A method for generating a stream topographical information using a drone and spatial information. The method according to claim 1,
The drones use fixed wing drones equipped with an airplane-shaped NiR camera,
A method of generating river terrain information using drones and spatial information, designed to shoot at a resolution of 6 cm at a flight altitude of 195 m, and a longitudinal and a lateral overlap of 85% and 70%, respectively. The method according to claim 1,
The step (a)
The images obtained through the camera shooting of the drone have basically WGS84 UTM coordinate system and 10 GCP (Ground Control Point) are selected for the land area of the river terrain for conversion to the GRS80 TM, which is a domestic geodetic coordinate system. A method for generating river terrain information using drone and spatial information, which performs VRS (Virtual Reference Station) survey for GCPs. The method according to claim 1,
Based on the point cloud data obtained through the camera shooting of the drones, a grading method is used to generate a DEM reflecting the elevation values available for stream surveying. In order to classify the trees in the stream using point cloud data, In order to extract the ground points, we first investigate the slope of the target area, calculate the slope using the distance between the point cloud data and the elevation difference based on the slope, and then calculate the slope of the ground And dividing the ground points into a plurality of point clouds, dividing cells having a predetermined size, dividing the points within a given inclination into ground points, and calculating an elevation among a plurality of point clouds existing in the cell after the virtual cells are generated We assign it to the lowest point seed pixel and calculate the slope with the neighboring cell How to Create Rivers terrain user is using, drones and space for applying a region growing method extends out of the cell within the threshold slope value specified. 5. The method of claim 4,
The region expansion method using point cloud data is a theory developed in the image field. Generally, pixels belonging to the same object region in an image are based on having similar statistical characteristics. Using these characteristics, a similar region is expanded by starting from an initial starting point, i.e., a seed pixel, by comparing with neighboring pixel values, and comparing the intensity values of neighboring pixels with the seed pixel, And,
The process of examining the homogeneity of regions in an image is examined by Equation (1)

(One)
Here, P (-) is called a logical predicate, and if True, the area in parentheses becomes the same area,

Denotes an initial seed pixel,

Is the number of neighboring pixels, and N is the total number of neighboring pixels. If the initial seed pixel and neighboring regions are homogeneous, then Equation (1) is satisfied. Expanded,
Based on the Microstation V8 and the Terra Scan / Terra Modeler, a digital elevation model (DEM) generation process based on the gradient technique was applied. The parameters applied in the preprocessing process depend on the density of the point clouds and the terrain conditions of the site The method of generating the stream topographical information using the drone and spatial information, which adjusts some values based on the parameter values commonly used in the existing studies and selects the DEM by visual reading. 6. The method of claim 5,
In the preprocessing step, all points are classified into a default class, and points having an elevation value that is too low or too high as compared with other points are designated as a low point class. Low point class classification is performed and the remaining default class The ground class is basically set by setting a maximum angle that can be generated, and each point is navigated through an inclination. In order to extract the ground class, a radius of 2 m The maximum distance that can be searched once was limited to 5m, and the process of finding the other points repeatedly based on the angle within the limited area was performed, and classification of ground points was performed. All points in the Default Class are classified as Low Vegetation Class and Low Vegetation Class When there is more than 0.5m books that are based on the fact that they exist as a Medium vegetation Class, more than 3m, and are classified as High vegetation Class,
In order to classify the misclassified points into the ground points by checking again from the orthoimage image, basically, since the ground is classified on the basis of 6 °, In order to prevent errors recognized as trees in the vicinity of the rivers, a terrain scan data processing software is used to modify the land surface by setting the area near the river bank as a separate area, Information generation method. The method according to claim 1,
As a result of analyzing the elevation distribution of the DSM data in the transverse section for the elevation values of the DSM and the DEM for the transverse section by selecting a part of the stream including the vegetation and the trees, In the case of DEM data generated by the gradient method based on the point cloud, the elevation distribution is higher than the elevation value of the ground A method for generating river terrain information using drones and spatial information, similar to The method according to claim 1,
100 verification points were selected to evaluate the accuracy of the DEM data generated through the DSM data obtained through the camera drone shooting and the point cloud data based gradient method. The verification points consisted of the terrain with little vegetation and the height A method for generating river terrain information using drone and spatial information, which selects various vegetation or trees to be included in various ways. The method according to claim 1,
The step (c)
VRS (Virtual Reference Station) survey and total station survey were performed for each verification point elevation. Horizontal position and elevation value were obtained from the VRS survey with little vegetation. A method of generating river terrain information using a drone and spatial information, which uses a total station to locate a prism on an adjacent terrain and then obtains an elevation value using a remote altitude surveying method because it is difficult to perform a VRS survey. 9. The method of claim 8,
Each elevation elevation for each verification point was obtained by performing VRS surveying and total station surveying. Horizontal position and elevation value were obtained through VRS (Virtual Reference Station) surveying of the terrain with little vegetation. , It is difficult to perform the VRS survey. Therefore, the total station survey was used to locate the prism on the adjacent terrain,
The results of the analysis show that the DEM generated from the point cloud data using the gradient method has a pattern similar to the maximum value (TRUE), whereas the DEM In the case of DSM, the method of generating river terrain information using drone and spatial information, which is similar to the maximum value (TRUE) at the verification point with little vegetation. 9. The method of claim 8,
The elevation error is calculated based on the maximum value (TRUE) obtained from the VRS survey and total station survey for DSM and DEM data by 100 verification points, and the height error range of the DSM based on the maximum value (TRUE) 0.08 ~ 15.33m, and the standard error was calculated as ± 3.84m. On the other hand, the height error range of DEM was 0.27 ~ 0.41m and the standard error was calculated as ± 0.12m. It was confirmed that the error was low,
In the case of stream survey where the elevation value of the actual ground should be reflected, the point cloud based gradient method is applied to the DEM from the DSM data obtained from the camera image of the drone (UAV)
A method for generating stream topographical information using Dron and spatial information, where DEM data analyzed by applying UAV point cloud - based gradient method is used as topographic data for performing river service in river maintenance basic plan, flood modeling, and sediment valuation.

Images