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-2/W25
出版社:Copernicus Publications
摘要:An automatic method of 3D textured building modelling with an airborne linear scanner sensor image and 3D model is proposed, which includes to extract buildings' feature from image, to match the extracted feature with 3D mode and to map texture according to the correspondence between image and model. Feature extraction from image and feature matching with 3D model are two key steps to build 3D textured building model automatically. In the article, a new line extraction approach through multi-level feature filter is presented, which is consisted of Canny edge detector, edge phase filter, edge direction filter with fault tolerance, Hough transformation, neighbour line segment fusion. Based line segmentation, a parallelogram extraction algorithm applying perceptual organization is proposed, which is based on perceptual organization, employing uncertainty reasoning and a shape expression form based on "part" to obtain simplicity, low computational overhead, and noise free. Matching between 2D image and 3D model is in essence to find a transformation matrix to minimize the error. A lot of algorithms are presented to solve the match problem with several unknown parameters, but still no good solution to problems with too much unknown parameters. Projected image will first be outputted based on camera model after partial matching between parallelogram extracted and 3D building. Then Hausdorff distance ia calculated between edge image and projected image. Based on which, feature matching is finished to extract texture and to map texture to buildings