Automatic vocabulary and graph verification for accurate loop closure detection

Localizing previously visited places during long-term localization and mapping, that is, loop closure detection (LCD), is a crucial technique to correct accumulated inconsistencies. In common bag-of-words (BoW) model, visual vocabulary built associate features for detecting loops. Currently, methods build vocabularies off-line determine scales of the by trial-and-error, which results in unreasonable feature association. Moreover, precision algorithm declines due perceptual aliasing given BoW-based method ignores positions features. To optimal automatically eliminate human heuristics, we propose natural convergence criterion based on comparison between radii nodes drifts descriptors construction. Furthermore, novel topological graph verification proposed validating candidates, can effectively distinguish ambiguities involving geometrical position thus improve LCD. Experiments various public datasets verify effectiveness our approach.