Automatic extraction of tree stem models from single terrestrial lidar scans in structurally heterogeneous forest environments

Abstract

An important application of terrestrial laser scanning is the extraction of tree stem models for diameter at breast height (DBH) assessment and forest inventory. Much work has been done to automate this process using adjacent co-registered lidar scans. Existing studies, however, have focused on pre-registered point clouds obtained from commercial lidar systems. We envision an affordable and efficient portable lidar hardware-software alternative for forest structural assessment. For such systems, point cloud registration software may not be available. Therefore we have developed automatic tree stem and DBH modelling approaches for single lidar scans collected from an off-the-shelf SICK LMS-151 lidar system. We assessed visual and quant itative accuracy for automatic tree extraction in a dense, heterogeneous, 20m x 20m forest site. We found that modelling tapered cylinders to the lidar point distribution of tree stems presents a viable algorithm for our single-scan lidar system, even in cluttered heterogeneous forest environments. This work demonstrates the potential for forest inventory using affordable, off-the-shelf lidar systems, and offers suggestions for future research. 1. In troduction Terrestrial laser scanning (light detection and ranging; lidar) has become an increasingly