Tree trunk detection system using LiDAR for a semi-autonomous tree felling robot

Abstract

To successfully fell a tree, semi-autonomous robotic devices must be capable of identifying the position, size and orientation of trees in the environment of a plantation forest. An approach is proposed to identify the aforementioned tree properties using data gathered from a Light Detection And Ranging (LiDAR) scanner. Trees are detected from the gathered point cloud by searching for horizontally directed normal vectors, then formed into clusters on which cylinder fitting is performed. This approach was verified using three different forest scenes to test its effectiveness.