Reconstruction of 3D building structures from IKONOS images through monoscopic line and shadow analysis

초록

Extraction of man-made objects from high resolution satellite images has been studied by many researchers. In order to reconstruct accurate 3D building structures most of previous approaches try to assume 3D information obtained by LIDAR or stereo analysis. They use high resolution images only to locate such 3D information in image space and define accurate building boundaries. However we argue that monoscopic image itself contains many clues of 3D information. When images are taken at oblique angles vertical lines of buildings can be extracted. Moreover shadows can explain heights of the corresponding objects. Since sun elevation and azimuth angles as well as sensor elevation and azimuth angles are mostly given in satellite metadata, it is possible to detect vertical lines and shadow lines automatically. Then by measuring the length of vertical lines or shadow lines we can estimate the height of buildings. On the other hands, we may determine building height interactively by projecting shadow regions for a given building height onto image space and by adjusting building height until the shadow region and actual shadow in the image match. Experiments with IKONOS images showed that the idea of monoscopic line and shadow analysis for extracting building height was feasible. The main paper will report the results of experiments together with quantitative assessment of performance.