期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2007
卷号：XXXVI-3/W52
页码：50-55
出版社：Copernicus Publications
摘要：Forest inventory schemes collect, besides tree species and some area parameters, geometric tree parameters such as diameter at breast height (DBH), tree height, stem profiles, azimuth and distance. For some years the use of a terrestrial laserscanner for this forestry inventory task has been discussed. Dense 3D point clouds recorded in forest stands may form the basis for automatic determination of forest inventory parameters. The paper presents an algorithm to detect trees in a horizontal cross section through a point cloud. This algorithm is divided in two segmentation steps to minimise the probability of false detections. The first segmentation step is a point cluster search in a cross section of the point cloud. In a second step all clusters are verified or discarded by analysing the point density in neighbouring cross sections. A study with 547 trees shows a detection rate of 97.4 % in single scan laserscanner data. Two other plots with heavy branching show a detection rate of 100 % and 94 %. Besides the tree detection, a new parameter is introduced to eliminate miss- fitted stem diameters. By using this parameter a least squares polynomial model is generated to smooth the diameters along the stem. Finally some results are demonstrated
关键词：Terrestrial laserscanning; point cloud; segmentation; automation; forest inventory; diameters along the stem
