Range image segmentation for tree detection in forest scans

Abstract. To make a tree-wise analysis inside a forest stand, the trees have to be identified. An interactive segmentation is often labourintensive and time-consuming. Therefore, an automatic detection process will aspired using a range image. This paper presents a method for the segmentation of range images extracted from terrestrial laser scanner point clouds of forest stands. After range image generation the segmentation is carried out with a connectivity analysis using the differences of the range values as homogeneity criterion. Subsequently, the tree detection is performed interactively by analysing one horizontal image line. When passing objects with a specific width, the object indicates a potential tree. By using the edge points of a segmented pixel group the tree position and diameter is calculated. Results from one test site are presented to show the performance of the method.