Random Decision Forests for Object Detection

Abstract

The image streams from the optical sensors in UAV (Unmanned Aerial Vehicle) are very large and highly dimensional, with considerable noise. Moreover, it is required to be capable of real-time information processing. In this paper we take advantage of random decision forests to learn a computationally efficient and accurate visual object detector for UAV. The random decision forests are learned with discriminative decision trees, where every tree internal node is a discriminative classifier. The experimental results show that our object detection approach achieves real-time performance and good object detection results.