Authors: Shichao Jin, Institute of Botany, Chinese Academy of Sciences, Qinghua Guo*, Institute of Botany, Chinese Academy of Sciences, Yanjun Su, Institute of Botany, Chinese Academy of Sciences, Tianyu Hu, Institute of Botany, Chinese Academy of Sciences, Hongcan Guan, Institute of Botany, Chinese Academy of Sciences
Topics: Remote Sensing, UAS / UAV, Environmental Science
Keywords: UAV, remote sensing, Lidar, forestry
Session Type: Paper
Presentation File: No File Uploaded
Accurately depicting the three-dimensional (3D) forest structure is the key to understand its functions and dynamics. Although in-situ and satellite remote sensing observations can provide rich large-scale forest observations, there is still a gap between the plot-scale measurements and the landscape/regional-scale measurements, especially the lack of 3D characterization of vegetation. Near-surface remote-sensing platforms, e.g. unmanned aerial vehicle (UAV) and terrestrial mobile system, have been rapidly growing due to their low cost, easiness to control, and high efficiency. The obtained very-high-resolution data can bridge the field measurements and satellite observations. In this study, we proposed a near-surface remote-sensing based solution for monitoring biodiversity in a new prospect. By integrating light-weight light detection and ranging (LiDAR) sensor, inertial measurement unit and global positioning system (GPS) together, we implemented a UAV-borne LiDAR system for capturing 3D ecosystem structures. A kinetic calibration system was developed to reduce the point cloud registration error from meters to less than 10 cm. Moreover, a backpack LiDAR system was built to collect under-canopy vegetation information in GPS-denial environment. A simultaneous localization and mapping algorithms were developed to automatically register LiDAR points in an environment without GPS signal. An automatic registration algorithm was further developed to fuse the UAV LiDAR and backpack LiDAR data based on individual tree locations, which can achieve an accuracy of 10 cm. The combination of these near-surface LiDAR systems has shown great capability to extract essential forest structure parameters (e.g., tree height, diameter at breast height) with high accuracy.