우선순위기반 특징선택과 시차맵을 이용한 보행자 검출

Abstract

In this paper, we proposed the pedestrian detection using priority-based feature selection and disparity map. In general, various features are used to improve pedestrian detection performance. However, using many features helps improve recognition performance, but it is disadvantageous in terms of speed. We only select features that have a great effect on pedestrian recognition, thus improving the speed without significantly degrading recognition performance. Through the experiment of ETH database, we verify the proposed method.