Comparing global measures of image similarity for use in topological localization of mobile robots

摘要

In this paper, two global measures of image similarity are compared for use in topological localization of mobile robots. A catadioptric vision system was used as the sensor. Our localization system employs a discrete Bayes filter. Experimental results with real images of an office environment indicate the robot can localize itself from the memorized training images even with the modification of the environment.