Point Based Matching Algorithm for Damaged Area Detection

Many disasters are frequently caused by earthquakes in Japan. In a disaster, the communication system is disrupted, and it is difficult to grasp the situation. Several remote sensing methods have been reported for detecting damaged areas in a disaster. Many methods use aerial photos taken after the disaster. Currently, satellite image libraries have already covered over 95 percent of the land area of Japan. If we use these images, we can obtain more information about the disaster. In this study, we use satellite images taken before the disaster and aerial images taken after the disaster, and we propose a point based matching algorithm that matches these two images for damaged area detection. We use this point based matching algorithm for reducing the difficulty of registering satellite images and aerial images. Satellite and aerial images are taken by different sensors and taken in different seasons. These cause false registration in some places. In our registration method, we use a selection algorithm for removing erroneous registration points. We remove false registration and use only well registered points for obtaining correct registration results. Then, we make the registered image by projective transformation using registered points. Finally, we detect damaged areas by comparing two images. For this purpose, we use the intensity of the red color of the images.