Multisource point cloud fusion for forest and post-fire forest mapping: Case study from the Bohemian Switzerland National Park
Abstrakt
Airborne and RPAS laser scanning and photogrammetry are well established techniques for forest mapping and monitoring. Nevertheless, they face challenges in complex, rugged terrain consisting of e.g., steep valleys or rock formations. Due to occlusion and canopy cover, LiDAR pulses do not reach the bare earth with sufficient density to reliably model the terrain and canopy height. Mobile, handheld terrestrial laser scanners can help to overcome this problem, though, from a practical point of view, they are best suited for smaller areas.
Our study is focused on the comparison and fusion of point clouds acquired with RPAS LiDAR, photogrammetry, and a handheld laser scanner. The collected dataset is quite unique in that it covers a relatively deep and narrow valley in a sandstone area that was originally forested but has now been significantly altered due to a recent forest fire. The main objective is to evaluate the suitability of each method for forest and terrain mapping in this specific area with respect to point cloud quality. In particular, the goals are to 1) assess absolute spatial accuracy of the point clouds based on check points, 2) evaluate relative accuracy/alignment between the point clouds, 3) test existing methods (ICP) for point cloud fusion to achieve the best possible dataset for terrain and forest modelling.