Ceiling feature detection for Geo-referencing in SLAM

In this study, we focused on ceiling features, such as illuminators, emergency sign boards, and Wi-Fi routers, to cancel accumulated errors in simultaneous localization and mapping. First, point cloud data in indoor spaces are acquired using a time-of-flight camera. Second, ceiling surfaces are estimated from the acquired point cloud data with the random sample consensus algorithm. Third, ceiling features are estimated to be used for reference features. Then, gravity points of the estimated features are estimated to be used for reference points. Finally, an indoor environment map is generated with the reference points. Through our experiments in indoor environments, we clarified that our methodology can detect ceiling features to be used for reference points.