Integration of visual and depth information for vehicle detection

Abstract — In this work an object class recognition method is presented. The method uses local image features and follows the part based detection approach. It fuses intensity and depth information in a probabilistic framework. The depth of each local feature is used to weight the probability of finding the object at a given scale. To train the system for an object class only a database of annotated with bounding boxes images is required, thus automatizing the extension of the system to different object classes. We apply our method in the problem of detecting vehicles captured from a moving platform. The experiments in a data-set of stereo images captured in an urban environment show a significant improvement in performance when using both information modalities. I