An ORB Feature Matching Algorithm for Mobile Devices

Feature matching is a key issue to realize a mobile-based augmented reality system. However, the performance of the matching methods is constrained by the processing speed and storage capacity of mobile devices. Considering the real-time requirement of tracking registration in the augmented reality system, a feature matching algorithm based on ORB is proposed, which combines SIFT algorithm with the original ORB algorithm to improve the weakness of the scale invariance on the basis of ensuring the efficiency of the algorithm. Moreover, a Bag-of-Words model is used to enhance the matching speed of feature points. Experimental results show that the algorithm can obtain better results than the original ORB algorithm on the aspects of real-time tracking target, scale invariance and matching time.