Visual place recognition using Bayesian Filtering with Markov Chains ∗

Abstract

Abstract. We present a novel idea to use Bayesian filtering in the case of place recognition. More precisely, our system combines global image characterization, Learned Vector Quantization, Markov chains and Bayesian filtering. The goal is to integrate several images seen by a robot during exploration of the environment and the dependency between them. We present our system and the new Bayesian filtering algorithm. Our system has been evaluated on a standard database and shows promising results. 1