On-line boosting-based car detection from aerial images

Car detection from aerial images has been studied for years. However, given a large-scale aerial image with typical car and background appearance variations, robust and efficient car detection is still a challenging problem. In this paper, we present a novel and robust framework for automatic car detection from aerial images. The main contribution is a new on-line boosting algorithm for efficient car detection from large-scale aerial images. Boosting with interactive on-line training allows the car detector to be trained and improved efficiently. After training, detection is performed by exhaustive search. For post processing, a mean shift clustering method is employed, improving the detection rate significantly. In contrast to related work, our framework does not rely on any priori knowledge of the image like a site-model or contextual information, but if necessary this information can be incorporated. An extensive set of experiments on high resolution aerial images using the new UltraCamD shows the superiority of our approach.