Monocular visual odometry for in-pipe inspection robot

This paper describes a monocular visual odometry technique for low texture environment localization. Estimating the pose of a robot in a small time interval is one of the challenging problems in robotics. Localization, mapping and motion planning are three fundamental problems which directly use the pose information of the robot to achieve their goal. In this work, we extract the pose information by processing image inside a concrete culvert which is taken by a camera attached to a mobile robot. We analyze different motion scenario using correlation based image to find displacement vector between two images. Real image data from the displacement of the robot are used to demonstrate the propose method. The experimental results show that the selected method proven efficient for the in-pipe inspection robot localization.