The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Download
Publications Copernicus
Download
Citation
Articles | Volume XLI-B3
https://doi.org/10.5194/isprs-archives-XLI-B3-411-2016
https://doi.org/10.5194/isprs-archives-XLI-B3-411-2016
09 Jun 2016
 | 09 Jun 2016

FAST AND ROBUST STEM RECONSTRUCTION IN COMPLEX ENVIRONMENTS USING TERRESTRIAL LASER SCANNING

D. Wang, M. Hollaus, E. Puttonen, and N. Pfeifer

Keywords: Terrestrial Laser Scanning, Stem Detection, Cylinder Fitting, RANSAC, Self-adaptive Cylinder Growing, Point Cloud Sampling

Abstract. Terrestrial Laser Scanning (TLS) is an effective tool in forest research and management. However, accurate estimation of tree parameters still remains challenging in complex forests. In this paper, we present a novel algorithm for stem modeling in complex environments. This method does not require accurate delineation of stem points from the original point cloud. The stem reconstruction features a self-adaptive cylinder growing scheme. This algorithm is tested for a landslide region in the federal state of Vorarlberg, Austria. The algorithm results are compared with field reference data, which show that our algorithm is able to accurately retrieve the diameter at breast height (DBH) with a root mean square error (RMSE) of ~1.9 cm. This algorithm is further facilitated by applying an advanced sampling technique. Different sampling rates are applied and tested. It is found that a sampling rate of 7.5% is already able to retain the stem fitting quality and simultaneously reduce the computation time significantly by ~88%.