Visual loop closure detection by matching binary visual features using locality sensitive hashing

In this paper, we present a novel approach for visual loop-closure detection in autonomous robot navigation. Our method uses locality sensitive hashing (LSH) as the basic technique for matching the binary visual features in the current view of a robot with the visual features in the robot appearance map. We show that this approach is highly efficient in comparison with using non-binary visual features such as SIFT and that it is more accurate than the popular bag-of-words (BoW) approach for generating loop closure candidates. Our experiment was conducted with an indoor dataset.