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:2009
卷号:XXXVIII-5/W1
出版社:Copernicus Publications
摘要:So far, realistic models of interiors have always been designed manually with the help of dedicated software packages. However, the demand for indoor models for different purposes has recently increased, thus a higher degree of automation could better satisfy different applications and speed up the processes. We present a technique for the fully automated modelling of indoor environments from a three dimensional point cloud. The results we achieve are very promising and the method suggested may provide completion to the actual standard for 3D city modelling. Our approach is based on a plane sweep algorithm for the segmentation of a point cloud in order to recognize the planar structures of a room. At first the 3D points that belong to the horizontal structures are tagged by sweeping a virtual plane along the vertical direction and thresholding the distances of each point to the plane. All the points that are not chosen as either .oor or ceiling are labelled as potential wall points and are being considered in the following segmentation step to detect the vertical faces. Finally, the .oor plan of the room is estimated by intersecting the directions of the walls and finding the vertices that constitute the ground shape. The result generated is a 3D model in CAD format, which perfectly fits the original point cloud
关键词:LIDAR; point cloud; segmentation; 3D model; plane sweep; floor plan