Comparison of UAV photograph-based and airborne lidar-based point clouds over forest from a forestry application perspective

TLDR

The test site is a forested area 15 km southeast of Jena, Germany. The study compares UAV photograph‑based point clouds and derived products with airborne lidar‑based data over this forested area. UAV data were collected over ~175 ha, with a 4 ha subset used for comparison at point‑like, surface, and tree‑level resolutions. The UAV data matched lidar products closely but provided finer detail, detecting 14 of 205 trees versus 45 missed by lidar, indicating superior small‑tree detection.

Abstract

In this study, unmanned aerial vehicle (UAV) image (photograph)-based point clouds and products were compared to airborne lidar-based data and products over a forested area. The test site is located in Germany, 15 km southeast of Jena. A total area of approximately 175 ha was covered during a UAV flight campaign. For this study, a subset of 4 ha (200 m × 200 m) was defined. The UAV–lidar comparison was accomplished at three different data levels: (1) point-like level (raster of maxima), (2) surface level (canopy height models), and (3) tree level (detection rate). In general, a high match between lidar- and UAV-based data/products was observed. The UAV data exhibits more details which are of particular importance for the detection of small trees. While using lidar data, 45 out of 205 trees were not detected, however only 14 trees were missed out when UAV data was used.