Vehicle detection from aerial images using local shape information

Abstract

Detection and extraction of vehicle objects in high resolution satellite imagery are required in many transportation applications. This paper presents an approach to automatic vehicle detection from aerial images. The initial extraction of candidate vehicle is based on Mean-shift algorithm with symmetric character of blob-like car structure. By fusing the density and the symmetry, the method can remove the ambiguous blobs and reduce the cost of the detected ROI processing in the subsequent stage. To verify the blob as a vehicle, log-polar shape descriptor is used for measuring similarity. The edge strengths are obtained and represented as its spatial histogram by the orientation and distance from the center of blob. The proposed algorithm is able to successfully detect the vehicle and very useful for the traffic surveillance system. © 2009 Springer Berlin Heidelberg.