Visual Place Recognition in Changing Environments with Sequence Representations on the Distance-Space Domain

Navigating in a perpetually changing world can provide the basis for numerous challenging autonomous robotic applications. With view to long-term autonomy, visual place recognition (vPR) systems should be able robustly operate under extreme appearance changes their environment. Typically, utilized data representations are heavily influenced by those changes, negatively affecting vPR performance. In this article, we propose sequence-based technique that decouples such from similarity estimation procedure. This is achieved remapping sequential representation into distance-space domain, i.e., domain which solely consider distances between image instances, and subsequently normalize them. way, perturbations related different environmental conditions embedded original vectors avoided, therefore scene efficacy enhanced. We evaluate our framework multiple with results indicating significant performance improvement over other approaches.