Tree segmentation and classification of deciduous park trees in Sanssouci Park, Potsdam, Germany, using airborne and terrestrial lidar point clouds

Abstract

<p>While automated, lidar-based tree delineation has proven successful for<br>conifer-dominated forests, deciduous tree stands remain a challenge.  But<br>automatic and reliable segmentation of trees at large spatial scales is a<br>prerequisite for a supervised classification into tree species. We propose an<br>aspect driven tree segmentation that clusters local elevation minima across<br>different aspects. These clusters define tree outlines that respect tree<br>inherent local elevation minima. We validate this approach with more than<br>25.000 mapped trees of the Sanssouci Park, Potsdam, using an airborne lidar<br>point cloud collected in 2018, and various terrestrial lidar scans for a large<br>fraction of the same park. Further, we demonstrate the tree segmentation by<br>supervised tree species classifications for the most common tree species using<br>random forests and Gaussian process classifiers with geometric parameters<br>derived from individual tree crowns.</p>