Semi-Automatic Building Reconstruction from Multi-Overlapping Aerial Images

• Main Authors: Guam-Hua Wang, 王冠華
• Other Authors: Liang-Chien Chen
• Format: Others
• Language: zh-TW
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/38470949860154664909

碩士 === 國立中央大學 === 土木工程研究所 === 91 === This investigation utilizes a semi-automated strategy to extract outlines for buildings from aerial images. First, we perform manual operation to locate the rough position for building corners in an image to obtain the approximate coordinates. Subsequently, accurate coordinates of those buildings may be reached by the proposed scheme. The procedures of the proposed scheme are described as follows. 1.The orientation parameters of the images are derived first. Then, rough initial locations for approximate building corners in the object space are given manually. The initial edges of those buildings are then projected to the image space. 2.We provide the approximate height of each building, then project to the overlapping aerial images. The predicted line segments for building outlines are then used to perform the feature detection. 3.An image matching procedure is applied for those conjugate line segments. 4.After performing space intersection, we calculate the 3-D line segments for buildings. The building model is, thus, generated. For reconstructing 3D model of buildings, the results show that the procedure we proposed is applicable for those buildings with good edge features.