Using 3D-Laser-Scanners and Image-Recognition for Volume-Based Single-Tree-Delineation and -Parameterization for 3D-GIS-Applications

Abstract

Today, 2D-GIS applications are standard tools in administration and management. As with all state-of-the-art systems certain environmental parameters were found to be difficult to perceive and convey, such that a novel 3D-presentation of the environment was developed to be more intuitive. The cost-effective graphics performance of current computers enables the move towards 3D-GIS and in turn, more detailed views of the environment. While current forestry-2D-GISs provide information on areas or collection of trees, the next generation of forestry-3D-GISs will store and make information available at level of an individual tree. In this paper we introduce an approach to bring vision technology into the forest. We present a volumetric algorithm based on the well-known watershed algorithm and use it to detect trees in laser-scanner point-clouds and four-channel aerial views. Based on this data, maps and virtual environments, ‘virtual forests’ are generated which can be used in a 3D-GIS for forest management, disaster management, forest machine navigation and other purposes.