Mobile robot loop closure detection using endpoint and line feature visual dictionary

Loop closure detection is a necessary component of mobile robot navigation. It can obviously reduce the accumulate error by detecting whether this place has been visited before. Environment information is gotten by cameras, because of it has the advantage of rich information, but how to represent place is still an open problem. Many methods based on BoW approach which extracts local feature points from scenes, and quantilized in visual word. This approach is fast and easy to implement, but suffers from perception aliasing, primarily due to vector quantization. In this paper, we propose to build endpoint and line visual dictionary which can describe endpoint set structure and texture information in environment, then two BoW vectors merged into one which contains relation of endpoint visual words. Experiment shows fusion feature is effective and better than single feature visual word, it shows a great improvement on recall at 100% precision when compared with state-of-the-art algorithms.