This paper proposes a new self-localization
algorithm, called dual-circle self-localization algorithm, for use in omnidirectional vision robots. When an autonomous robot navigates in the field, it senses the environment by an omnidirectional vision device fitted on the robot. The landmarks information is acquired by using image processing and pattern recognition techniques. Robot position on the field can be estimated with the proposed method by using these landmarks information. Comparing with other self-localization methods, the dual-circle self-localization algorithm requires only three landmarks information in the image for identifying the robot position, which is very efficient for image processing. Practical experiments on an autonomous robot have demonstrated that the proposed algorithm can achieve a satisfactory
performance to locate the robot.
Journal of the Chinese Institute of Engineers 35(6), pp.619-631