Shenglan Du

Accurate, detailed and automatic tree modelling from point clouds

Trees are of great significance throughout the world, and their models are widely applied in varies fields, such as landscape design or geo-simulation. Recent developments in laser scanning technologies make it possible to effectively acquire geometric attributes of trees and achieve accurate 3-dimensional tree modelling. In this thesis, a novel method is proposed to reconstruct tree branches accurately and automatically from laser scanned points. A Minimum Spanning Tree algorithm is employed to extract an initial tree skeleton which is then further simplified through iterative removal of redundant components. A global-optimization approach is performed to fit a sequence of cylinders approximating the geometry of the tree branches. The approach sows to be adaptable to various trees with different data qualities. Topological fidelity and geometrical accuracy are achieved without significant user interactions. The resulting tree models can be further used in the precise estimation of tree attributes, urban landscape visualization and other related applications.