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：2002
卷号：XXXIV-3/W4
出版社：Copernicus Publications
摘要：Lidar topographic surveys of forested terrain generate XYZ positions for laser returns from numerous points, some on the ground and some from vegetation. Extracting a ground surface model from such data requires 'virtual deforestation' (VDF), preferably by automatic means. A simple error budget for lidar topography of forested terrain suggests that the dominant source of error—and the greatest room for improvement—lies in VDF procedures. W e discuss a despike VDF algorithm that classifies returns as ground or not-ground on the basis of the geometry of the surface in the neighborhood of each return. The despike algorithm is fully automatic, effective, and can recover breaklines. It fails to identify some negative blunders, rounds some sharp corners off the landscape, and as implemented is slow. There are clear paths to improve its speed. If multiple-return data are available, a no-multiple-returns VDF algorithm robustly defines areas where all returns are ground returns. Many groups are using variations on block-minimum VDF algorithms, but these do not work well on slopes and typically require sub- stantial human involvement to adjust block size as the fraction of ground returns changes. Fully automatic VDF algorithms are desirable not only to minimize survey costs but also to produce topography for which all necessary interpretive biases and assumptions are explicit. The development of effective VDF algorithms has been hindered by the tendency of some commercial and academic practitioners to keep their work proprietary. Open dialogue is needed
