Simulation of LiDAR data for forestry applications

Abstract: In forestry it is important to have accurate information about the forest. LiDAR (laser scanning) can be used to scan vast areas of forest and from the data extract information about the trees. The purpose of this thesis is to develop a simulator for LiDAR data. The simulator will be tested on a method for tree localization (Holmgren and Lindberg 2013) to see how parameters like tree density and laser frequency effects the accuracy of the localization. First a simulator which uses simple shaped trees (in the shape of cones) is written. Later on a tree model based on real laser data is created by the use of histogram density estimation. Ray-tracing is used to simulate the LiDAR data which the trees give rise to. This is done by following each ray of laser and see where it is reflected. The tree localization method is tested on the data and we report the following findings: 1: The percentage of correctly located trees decreases with increasing tree density. 2: Larger trees yields an increase in false trees found by the localization method. 3: Higher laser pulse density decreases the number of false trees. 4: The minimum radius at which the localization method start fitting ellipsoids greatly effects the number of false trees. Smaller radius yield more false trees.