Digitizing notes using a moving smartphone : Evaluating Oriented FAST and Rotated BRIEF (ORB)

Abstract: This thesis investigates the problem of tracking objects for an Augmented Reality (AR) setting. More specifically, the issue of tracking Post-It® notes to be used in a Mobile Augmented Reality (MAR) application using the Oriented FAST and Rotated BRIEF (ORB) keypoint extractor and descriptor, is investigated. This problem explores the relatively new and unexplored territory of tracking specific objects in real-time on mobile devices. Since MAR is becoming more prevalent, this is a field that is likely to be explored in more depth in the future. A solution was implemented in an existing note scanning application. Test sequences, with accompanying ground truth, were created for the applicable scenarios. The test sequences were used to reliably verify and evaluate the implementation with regard to precision, recall, accuracy, and speed. The ground truth was generated in a Mixed-Initiative Computing (MIC) application. The results show that tracking using only ORB is not viable if high precision, recall, or accuracy is needed. While tracking via ORB may not be viable as a standalone solution, the thesis describes methods for using it in a MIC setting, which may be viable.