On-line Boosting for Car Detection from Aerial Images

In this paper, we present a new approach for automatic car detection from aerial images. The system exploits a robust machine learning method known as boosting for efficient car detection from high resolution aerial images. We propose to use on-line boosting with interactive training framework to efficiently train and improve the detector. We use integral images for fast computation of features. This also allows to perform exhaustive search for detection of cars after training. For post processing, we employ a mean shift clustering method, which improves 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.