3D information retrieval for visual odometry system of planetary exploration rovers - A stereo vision approach

Determination of 3-D data from the images is important in the field of machine vision. The most direct way to achieve this is stereo vision. 3-D data obtained from stereo images form the basis for many higher level tasks like robot navigation, object recognition, 3-D modeling, path planning etc. in the field of interplanetary mission, hazardous environments, deep sea exploration, autonomous car driving. In this work, we attempt to recover the 3-D information from the stereo images for developing the vision based navigation of future planetary rover missions. This paper details the process flow and algorithm for the retrieval of 3-D information from the images obtained from our stereo rig. The results obtained in this work closely matches with ground truth.