IV-WA: An Information Visual Tool with WebAR

Early marker-based mobile augmented reality technology relies on special patterns and cannot be applied to tracking natural feature objects. At present, more and more applications are prone to track natural feature objects, which gradually advances the natural features-based development of mobile augmented reality. However, current mobile augmented reality applications depend heavily on open-source software development kits like AR- Toolkit which only recognize a single object rather than multiple natural feature objects. Inspired by employing a vocabulary tree model to retrieve query sentences from an article, this paper proposes a multiple targets retrieval strategy called Pruning Inverse Visual Vocabulary Tree (PIVVT), which converts the feature matching into an N-ary tree searching feeding the real-time. Due to the instability of traditional vocabulary tree construction, which leads to insufficient storage, low retrieval speed, and accuracy, we reverse tree construction to fill up the entire tree from the bottom to the top to forms a full binary tree with the least amount of the storage, which builds a stable vocabulary tree. Finally, the test results prove that the strategy can improve the speed and accuracy of multiple natural feature target retrieval in mobile augmented reality.