Robust and low complexity obstacle detection and tracking

Abstract

Obstacle detection and tracking is essential module for autonomous driving. Vision based obstacle detection and tracking faces huge challenges due to factors like cluttered background, partial occlusion, inconsistent illumination, etc. In this paper, we propose a robust and low complexity stereovision based obstacle detection and tracking framework. Low complexity techniques are employed to detect obstacles in the u-v-disparity image space. In addition, effective strategies are proposed to construct a distinctive object appearance model for data association efficiently. Finally, an online multi-object tracking framework is proposed by integrating the obstacle detection and data association modules in a robust way. Extensive experimental results on the well-known KITTI tracking dataset demonstrate that the proposed method is able to detect and track various obstacles robustly and efficiently in diverse challenging scenarios.