Skeleton-based botanic tree diameter estimation from dense LiDAR data

Abstract

New airborne LiDAR (Light Detection and Ranging) measurement systems, like the FLI-MAP 400 System, make it possible to obtain high density data containing far more information about single objects, like trees, than traditional airborne laser systems. Therefore, it becomes feasible to analyze geometric properties of trees on the individual object level. In this paper a new 3-step strategy is presented to calculate the stem diameter of individual natural trees at 1.3m height, the so-called breast height diameter, which is an important parameter for forest inventory and flooding simulations. Currently, breast height diameter estimates are not obtained from direct measurements, but are derived using species dependent allometric constraints. Our strategy involves three independent steps: 1. Delineation of the individual trees as represented by the LiDAR data, 2. Skeletonization of the single trees, and 3. Determination of the breast height diameter computing the distance of a suited subset of LiDAR points to the local skeleton. The use of a recently developed skeletonization algorithm based on graph-reduction is the key to the breast height measurement. A set of four relevant test cases is presented and validated against hand measurements. It is shown that the new 3-step approach automatically derives breast height diameters deviating only 10% from hand measurements in four test cases. The potential of the introduced method in practice is demonstrated on the fully automatic analysis of a LiDAR data set representing a patch of forest consisting of 49 individual trees.