趙鍵哲2006-07-252018-07-092006-07-252018-07-092005http://ntur.lib.ntu.edu.tw//handle/246246/2902The authors investigated in this study by exploiting the concepts of structuralization on LIDAR point cloud and registering laser scanning data sets from different stations. The former task conducted by two streams includes extracting planes as well as line features and point features derivation. One of the two methods is to table the LIDAR point cloud attributed with geometric and topologic relationship of the points within near neighborhood and thus be able to classify the point cloud into planar features. Line features and point features are then derived seeking neighboring planes. The other method is to employ 3D grid structure well addressing point cloud into 3-D topology and followed by (a). region growing for hypothesizing planes and (b). iterated Hough Transform for refining the plane-features. Line features and point features can then be derived based on the previous solution on plane extraction. The latter task is mainly focused on developing the algorithms in which 3D line feature correspondences for registering multiple data sets with overlapping scene can be established by using geometric constraints in an automatic fashion. The spatial similarity transformation can then be performed by matched 3D line feature correspondence.application/pdf3615380 bytesapplication/pdfzh-TW國立臺灣大學土木工程學系暨研究所StructuralizationRegion GrowingIterated Hough Transform3D Line FeaturesSpatial Similarity Transformationreporthttp://ntur.lib.ntu.edu.tw/bitstream/246246/2902/1/932211E002054.pdf