A study of visual odometry for mobile robots using omnidirectional camera

摘要

This paper describes a new visual odometry system using omnidirectional images for application to mobile robots under an urban environment. Both velocity and direction of movement can be estimated from a sequence of omnidirectional images. A unique Hough transform based algorithm enables real-time estimation of the speed of a moving robot. The validity of the proposed method is examined using a wheelchair-type mobile robot.