4:15 PM - 4:30 PM
[HTT14-04] Comparative study of laser and photogrammetric surveying using multi-rotor UAV in coniferous forest zones
Keywords:aerial survey, SfM, LiDAR, point cloud data, UAV, Mt. Fuji
In the coniferous forest zone at the northern foot of Mt. Fuji, we conducted laser and photogrammetric surveys using a multi-rotor UAV to obtain digital surface model including vegetation. For the laser survey, we used LiDAR, GreenValley International LiAir X3C-H, which has a wavelength of 905 nm, 32 channels, FOV: 360° for horizontal × 40.3° for vertical, and can scan up to 300 m at 640,000 points/sec. It has a built-in 26MP mapping camera. For the photogrammetry, we used DJI Zenmuse H20T, which is equipped with three cameras: a telephoto camera, a wide-angle camera, and a thermal camera. The photos taken with the wide-angle camera (1/2.3-inch CMOS, 12MP) were used for structure from motion (SfM). The survey area is a coniferous forest zone at an altitude of about 1,000 m at the northern foot of Mt. Fuji, where mainly red pine trees grow at heights of 30 to 40 m. The survey area is a square with sides of approximately 100 m, and has an area of approximately 10,900m2. The LiAir or H20T was mounted on DJI Matrice 300 RTK, an industrial multi-rotor UAV, and aerial surveys were performed under the following conditions. The altitudes above ground are 50 m or 100 m, the flight speed is 1 m/s, the side lap ratio is 70%, and the front-back overlap ratio is 90%. GreenValley LiDAR 360, a point cloud analysis software, was used to generate cloud point data acquired by the LiDAR. Orthomosaic images and colored point cloud data were generated from the aerial photos using the photometric software PIX4Dmappar. These point cloud data were visualized and compared using the open-source point cloud processing software CloudCompare.
Firstly, the data acquired by LiDAR surveying at an altitude of 50 m had the highest point cloud density, and the reproduction of fine details such as tree branches was the best. On the other hand, the photogrammetric point cloud data exhibits lower density and less accurate tree representation compared to the LiDAR data. However, the data were higher than the point cloud data for the whole of Yamanashi Prefecture generated by aerial laser surveying using manned aircraft (Yamanashi Prefecture, Department of Land Development, Fuji-Toubu Construction Office). In this way, laser surveying and photogrammetry using muti-rotor UAVs, which fly at low-altitude and slow speeds, are practical for acquiring surface data of geographical features, including vegetation. However, the number of point clouds from the ground to near tree height is limited. To compensate for this, it is necessary to conduct surveys using, for example, a handheld LiDAR. On the other hand, aerial laser surveying using manned aircraft that fly at high altitudes and high speeds tends to produce point cloud data with relatively low point density. However, they can survey a wide area than multi-rotor UAVs, making them suitable for wide-area terrain analysis. For example, upon checking the wide-area DEM generated from the point cloud data obtained by the aerial surveys using manned aircraft, we can see that the surveyed area is located on a lava flow terrain with a width of approximately 500 m. That is to say, when performing terrain analysis, it is important to choose and/or combine the surveying methods appropriately according to the target, area, required accuracy, and budget, etc.
Firstly, the data acquired by LiDAR surveying at an altitude of 50 m had the highest point cloud density, and the reproduction of fine details such as tree branches was the best. On the other hand, the photogrammetric point cloud data exhibits lower density and less accurate tree representation compared to the LiDAR data. However, the data were higher than the point cloud data for the whole of Yamanashi Prefecture generated by aerial laser surveying using manned aircraft (Yamanashi Prefecture, Department of Land Development, Fuji-Toubu Construction Office). In this way, laser surveying and photogrammetry using muti-rotor UAVs, which fly at low-altitude and slow speeds, are practical for acquiring surface data of geographical features, including vegetation. However, the number of point clouds from the ground to near tree height is limited. To compensate for this, it is necessary to conduct surveys using, for example, a handheld LiDAR. On the other hand, aerial laser surveying using manned aircraft that fly at high altitudes and high speeds tends to produce point cloud data with relatively low point density. However, they can survey a wide area than multi-rotor UAVs, making them suitable for wide-area terrain analysis. For example, upon checking the wide-area DEM generated from the point cloud data obtained by the aerial surveys using manned aircraft, we can see that the surveyed area is located on a lava flow terrain with a width of approximately 500 m. That is to say, when performing terrain analysis, it is important to choose and/or combine the surveying methods appropriately according to the target, area, required accuracy, and budget, etc.