단일 어안 카메라 기반 후방 장애물 높이 추정 방법

Abstract

Generally, Stixel is based on a stereo camera, and used for the forward environment representation. This paper proposes monocular camera-based Stixel for the rear environment representation. In particular, this paper focuses on the obstacle height estimation using sparse disparity map. Disparity and appearance information is fused in this paper since the sparse disparity map acquired by motion stereo has little information compared to dense disparity map. Experimental results show that the proposed method can successfully estimate obstacle height in actual parking lots.