As an effective means for acquisition of three-dimensional scene, Terrestrial Laser Scanning (TLS) has been widely used in surveying and mapping of historic buildings. However, the application of TLS to large-scale building surveying and mapping projects is still faced with the reality of inefficiency. Based on the experience gained in the theoretical study and practice of building surveying and mapping, the author of this thesis attribute factors that influence the efficiency of building surveying and mapping to three bottlenecks:point clouds registration, point cloud recognition and filtering, as well as generation of three-view drawing of buildings, which are heavily dependent on human-computer interaction. Around these three key problems, the author has carried out related research. In the aspect of automatic registration of multi-scan point clouds, the author has proposed automatic registration of multi-scan point clouds using precisely located targets. In the aspect of point cloud recognition and filtering, the authors has proposed color-based segmentation method for point clouds and point cloud filtering based on scanning point statistics. And the smoothness constraint point cloud segmentation as well as point cloud classification based on neighborhood Principal Component Analysis (PCA) have been implemented. In the aspect of generation of three-view drawing of buildings, the author has proposed automatic extraction of architectural feature lines in orthographic depth images. In addition, in the aspect of feature point detection, the author has improved method for corner detection in the open source point cloud processing software Point Cloud Library (PCL). The research results of this thesis can be used to effectively improve the efficiency of large-scale building surveying and mapping, in which TLS is the main means of data acquisition. At the same time, the proposed methods and techniques will also help addressing related problems in the field of photogrammetry and computer vision. |