Abstract
A novel method for visual place recognition is intro-duced and evaluated, demonstrating robustness to percep-tual aliasing and observation noise. This is achieved byincreasing discrimination through a more structured repre-sentation of visual observations. Estimation of observationlikelihoods are based on graph kernel formulations, uti-lizing both the structural and visual information encodedin covisibility graphs. The proposed probabilistic modelis able to circumvent the typically difficult and expensiveposterior normalization procedure by exploiting the infor-mation available in visual observations. Furthermore, the place recognition complexity is independent of the size of the map. Results show improvements over the state-of-theart on a diverse set of both public datasets and novel experiments, highlighting the benefit of the approach.