Automatic tree detection and attribute characterization using portable terrestrial lidar

Abstract

Key messageThis study details a methodology to automatically detect the positions of and dasometric information about individual Eucalyptus trees from a point cloud acquired with a portable LiDAR system.Currently, the implementation of portable laser scanners (PLS) in forest inventories is being studied, since they allow for significantly reduced field-work time and costs when compared to the traditional inventory methods and other LiDAR systems. However, it has been shown that their operability and efficiency are dependent upon the species assessed, and therefore, there is a need for more research assessing different types of stands and species. Additionally, a few studies have been conducted in Eucalyptus stands, one of the tree genus that is most commonly planted around the world. In this study, a PLS system was tested in a Eucalyptus globulus stand to obtain different metrics of individual trees. An automatic methodology to obtain inventory data (individual tree positions, DBH, diameter at different heights, and height of individual trees) was developed using public domain software. The results were compared to results obtained with a static terrestrial laser scanner (TLS). The methodology was able to identify 100% of the trees present in the stand in both the PLS and TLS point clouds. For the PLS point cloud, the RMSE of the DBH obtained was 0.0716, and for the TLS point cloud, it was 0.176. The RMSE for height for the PLS point cloud was 3.415 m, while for the PLS point cloud, it was 10.712 m. This study demonstrates the applicability of PLS systems for the estimation of the metrics of individual trees in adult Eucalyptus globulus stands.