Monocular height estimation by chronological correction of road unevenness

Abstract

Height estimation of obstacles is a valuable resource for safer and accurate locomotion of autonomous robots and vehicles. This research proposes a new height estimation method using only a monocular camera based on chronological correction of road unevenness. The method applies the traditional flat surface model to compute the distance to points on the ground but estimates and corrects the influence of unevenness in each frame. In the conducted experiments, the proposed height estimation method could achieve an average error of 41.8 mm, which is a smaller error than existing techniques.