VISUAL ODOMETRY TECHNIQUE USING CIRCULAR MARKER IDENTIFICATION FOR MOTION PARAMETER ESTIMATION

摘要

This paper presents a new visual odometry approach for mobile robot self-localization utilizing natural circular invariant features during motion. It is proposed that the onboard camera acquires sequences of overlapping mages and senses the distance and orientation of the vehicle with respect to identified markers. The paper uses an effective image filtering technique based on convolution that can be used to localize the natural markers in the images. The proposed approach simplifies the problem of feature localization and allows a robust estimation of the vehicle's trajectory. Initial experiments are carried out on a mobile robot and results are presented.