Monocular vision based independently moving feature detection using image correspondences

摘要

In this paper, we propose a independently moving feature detection algorithm using monocular vision. The proposed algorithm can detect independently moving feature points from all of the feature points by geometric constraints. The geometric constraints consist of epipolar constraint and trifocal constraint. The proposed algorithm can be implemented in real time. Therefore, it can be directly apply to mobile robots and intelligent vehicles. We test our algorithm with real outdoor datasets.