Detecting building changes using epipolar constraint from aerial images taken at different positions

A new method for detecting building changes from aerial images taken at different positions and times is presented. In these images, various changes are observed simultaneously. The method detects building changes by distinguishing them from the image differences caused by parallax and shadow. First, the roofs of buildings are identified and corresponding regions on epipolar lines, which are derived from the epipolar constraint, are identified. The regions where correspondency fails are taken as building change candidates. Second, shadow effects are eliminated by analyzing the illuminance levels, i.e., changes in image intensity levels. Experiments using actual images show that the proposed method can well detect building changes