Extraction of road traffic information using satellite images and a three-dimensional digital map

Analysis of traffic information is one of the applications of remote sensing. Several studies have been reported for vehicle extraction from satellite images or aerial images by using image processing methods. The analysis of these images is not influenced by the ground damage and can obtain a lot of information over a wide area. In such studies, the shadow areas casted by buildings are the cause of errors in extracting vehicles in urban areas. This is because the shadow areas are dark and the positions of vehicles in the areas are unclear. In this paper, we propose a method of extracting shadow areas casted by buildings using three-dimensional digital map data of buildings and extracting vehicles in the areas using image processing methods. The conventional method of extracting shadow areas uses the image intensity, however, this method has the problem that objects having low intensity are mis-extracted. Our method solves this problem by estimating the position and shape of shadow areas by using three-dimensional digital map data and metadata of a satellite image. In vehicle extraction, we use edge detection method for detecting the outlines of vehicles. The detection of the vehicle edges is difficult, since the intensities of vehicle edges are different in the sunny areas and in the shadow areas. However, by extracting shadow areas using the map data in advance and computing the threshold of the edge detection dynamically, our method can detect the vehicle edges and obtain the vehicle positions correctly. We developed relevant software on the computer, and we analyzed actual images to evaluate the effectiveness of our method.