A Comparative Study of Building Edge Extraction on Aerial Images

• Main Authors: Sen-Che Tsao, 曹森哲
• Other Authors: Jung-Der Tsai
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/23007333918788143964

碩士 === 國立中興大學 === 土木工程學系所 === 97 === This research studies the extraction of boundary features of buildings from aerial images. Traditionally, the accuracy of manual photogrammetric measurements of linear boundaries is subjective to the operator’s conscience and skill and is usually hard to evaluate. Therefore, the purpose of this research is to study automatic methods on edge extraction as alternative for the time-consuming manual measurements. In this research, several edge detectors, including Canny, Sobel, Prewitt, Robert and Laplacian of Gaussian (LoG), were applied on the experimental aerial images for extracting the edge pixels, followed by edge linking process for grouping straight line segments. The intersection of these line segments were then computed as the corners of corresponding buildings for comparison with manual measured results. Experimental results from this research show that the automatic method on building corner detection is much easier than manual approach and could provide much accurate information through appropriate setting of threshold on edge detection.