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：2005
卷号：XXXVI-5/W17
出版社：Copernicus Publications
摘要：This paper presents a general strategy for 3D reconstruction of complex architectural scenes. This approach makes use of both laser data (range and retro-diffusion) and image data, and of their great complementarity. The proposed approach starts with pose estimation between the two data sets. This part is largely automated by means of correlation of fine feature points and linear primitives. The second part deals with segmentation of the scene. Here, each element of the scene must be detected. This can be fulfilled in two ways. The first one is semi-automatic extraction of the objects where user interaction is used to initialize the object localisation, the second one is fully automatic segmentation, driven by image and 3D data. Then, each object needs to be reconstructed. We consider a large range of objects since we can describe planar surfaces, general geometric primitives and generalised cylinders. When these models do not fit with the surface, that is to say on complex parts, triangulation is performed. This leads to two potential reconstruction tools: one similar to laser data editing software, with more efficient object selection and object boundary estimation, the other would perform automatic reconstruction, which would be followed by a verification procedure. The advantage of this approach, compared to existing techniques, is to reduce dramatically reconstruction time, without having any hard limitation on the set of shapes that can be obtained
关键词：Laser scanner; image data; terrestrial surveying; 3D reconstruction; semi-automation; data fusion
