An image matching method based on BoVW model for visual self-localization

An image matching method based on the bag of visual words model is presented, and applied to visual self-localization of mobile body. The local feature of image extracted by SURF is employed to produce the bag of visual words for every image. “From coarse to fine” double-layers indexing structure is established based on the grouped image set. The approach of hierarchical-clustering is used to generate the bag of visual words for image group. Double-layers indexing structure reduces the searching space of image matching, the memory load and the computational complexity sharply. Based on the double-layers indexing structure, the mobile body visual localization includes two steps: rough location and accurate location. The rough location based on group index is fast, but not accurate. As for the accurate location, it is based on the inner group index, and is relatively slow, but has high accuracy. The integration of the above two methods can improve the speed and precision. Outdoor scene experiment results show the effectiveness and the feasibility of this method.