AUTOMATIC TREE DETECTION/LOCALIZATION IN URBAN FOREST USING TERRESTRIAL LIDAR DATA

Abstract

Individual tree detection is essential task to access relevant parameters at the tree scale, such as: diameter at breast height (DBH), first branch height, and tree height. In this context, we propose an automatic tree detection/localization approach based on trunk geometry, i.e., on vertical continuity, not requiring preprocessing stages (ground filtering, point cloud normalization, classification) or training samples, as in some classes of methods. The performance of the proposed approach was evaluated using LiDAR data acquired by a terrestrial laser scanning (TLS) system in an urban forest. Obtained results indicated the potential of the proposed approach, resulting in an Fscore of 98% and a RMSEXY of 15 cm.