A Study on Quality Refinement of Semi-Automatic Building Mapping Using Digital Aerial Images

• Main Authors: Sheng-Yi Wang, 王勝益
• Other Authors: Victor J. D. Tsai
• Format: Others
• Language: zh-TW
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/63187961661098404616

碩士 === 國立中興大學 === 土木工程學系所 === 94 === A stereo pair of two overlapped aerial images was usually used in traditional manual stereo compilation for determining the 3-D ground coordinates of topographic features. The quality of such digital compilation without redundancy depends on the experience and skill of the operator and is usually examined by ground checking of limited samples. The information residing in multiple overlapped images thus lies waste. This research develops a semi-automatic approach in quality refinement for digital compilation. The approach uses manual measurements of buildings from a stereo pair of aerial images as initial approximations in applying edge extraction, Hough transform, noise removal, line detection with additional constraints to extract the image coordinates of the building corners from the multi-overlap images. These image coordinates are then used in the space intersection for a least squares solutions of the spatial coordinates of the buildings and in order to promote the precision of detail mapping. Experimental results from using multi-overlap aerial images in 1/1,000 and 1/5,000 topographic mapping projects demonstrate that precise corners can be satisfactorily obtained for most buildings by applying the proposed approach in this research. So is the quality refinement for efficient assessments on manual measurements.