Co-host: The Institute of Image Electronics Engineers of Japan, The Institute of Image Information and Television Engineers
This paper proposes a method to automatically extract building areas and to distinguish the roof shape of each building from a high resolution aerial photograph. First, all the straight lines are extracted from an aerial photograph by Hough transformation, and building areas are extracted by detecting straight lines parallel to the roads since the boundary of the building region is usually composed of straight lines, and is parallel to the road. Next, edges included in every building area are extracted with Sobel operaton. Moreover, the brightness difference among roof surfaces caused by the sunlight is examined. As a result, the extraction rate of building areas was only 58.6%. However, the distinction rate of the roof shape against every building area has reached 87.9% by the edge extraction.