Comparison of Point Cloud Data Reduction Methods in Single-Scan TLS for Finding Tree Stems in Forest

The point density in a single-scan terrestrial laser scanner (TLS) point cloud is very dense close to the scanner and gets sparser as the distance from the scanner increases. A full circular scan can contain tens of millions of points, which is impractical for most algorithms that work on point data. The number of points can be reduced by taking a sample of the original data. We have studied what influence different sampling methods have on the number of points that falls on tree stems. We propose that the number of points available on a far-away tree can be increased with a smart data reduction scheme. The data reduction favours far-away points over the densely located points close to the scanner. The main findings of this study are that removing ground points before sampling gives a great advantage in data reduction and that a point selection using only horizontal distances (2D Cartesian, xy-plane) favours low points.