Stereoscopic mobile Natural Feature Tracking
Rochester Davies, Ian David Wilson, Andrew Ware · 2014
In this paper, we introduce a Natural Feature Tracking (NFT) system capable of running at real-time on limited powered devices. Combining current vision techniques with stereoscopic cameras has benefits that reduce processing expense and increase the suitability of the resultant data. The novel use of stereoscopic imaging centered within a single device allows for extraction of features from dynamic backgrounds where the camera is not bound to a single fixed point and is subject to 6DOF. The System achieves frame rates in excess of 15fps operating on a single 1GHz core mobile device. The device used has a dual core allowing the addition of virtual content for augmentation without affecting the performance of the tracker.