Visual odometry for robust rover navigation by binocular stereo

Abstract

Abstract: This work deals with the problem of estimating the trajectory of an autonomous rover by passive stereo vision only (visual odometry). The proposed method relies on the tracking of pointwise image features and on the estimation of the robot motion by robust bundle adjustment of the stereo matched features, before and after the motion. Preliminary results from the application of the algorithm both to simulated data and to actual image sequences acquired by a mobile platform are presented and discussed.