Sparse Visual Localization in GPS-Denied Indoor Environments

This paper presents a real time feature based sparse visual localization algorithm for indoor environments such as shopping mall. ORB feature is used for feature detection and descriptor computation in this project. The features extracted from ORB are mainly points in corners, and edges. Since this project is used for indoor localization, the dataset for experiment is TUM RGBD dataset. This dataset contains various indoor environments which can fulfill our requirements. The essential and fundamental matrices are solved based on the matching of ORB features and Eight-point algorithm. We show certain advantages of sparse visual localization algorithms in our experiment's results.