| Title |
Entropy-based pre-filtering of high-density forest point clouds for individual tree detection algorithms |
| Authors |
Karlauskas, Kasparas ; Treigys, Povilas |
| DOI |
10.1145/3748777.3748806 |
| ISBN |
9798400720949 |
| Full Text |
|
| Is Part of |
SSTD'25: Proceedings of the 19th international symposium on spatial and temporal data, Osaka, Japan August 25 - 27, 2025.. New York : Association for Computing Machinery, 2025. p. 137-140.. ISBN 9798400720949 |
| Keywords [eng] |
forestry ; tree stem segmentation ; individual tree detection ; LiDAR ; entropy |
| Abstract [eng] |
Light Detection and Ranging have enabled the creation and analysis of highly detailed 3D point cloud representations of outdoor areas through Terrestrial Laser Scanning and Mobile Laser Scanning. In the field of forestry, one important spatial data processing task is detecting individual tree stems, which is often a prerequisite for individual tree location mapping, wood volume estimation, and other tree inventorization analysis. The high point densities achieved through Terrestrial Laser Scanning and Mobile Laser Scanning allow for high accuracy in delineating individual tree stems, but lead to a high computational cost due to the large point counts. This study proposes a computationally efficient approach for removing points unlikely to belong to tree stems, based on point distribution entropy within local regions. The approach is motivated by the observation that stems exhibit low entropy, whereas canopy regions tend to have higher entropy. The point filtering algorithm works in linear time and can lead to significant processing time gains while often maintaining accuracy for downstream digital forestry tasks. |
| Published |
New York : Association for Computing Machinery, 2025 |
| Type |
Conference paper |
| Language |
English |
| Publication date |
2025 |
| CC license |
|
