Applying Saliency-Based Region of Interest Detection in Developing a Collaborative Active Learning System with Augmented Reality

摘要: Learning activities are not necessary to be only in traditional physical classrooms but can also set up virtual environment. Therefore the authors propose a novel augmented reality system organize class supporting real-time collaboration and active interaction between educators learners. A pre-processing phase is integrated into visual search engine, heart of our system, recognize printed materials with low computational cost high accuracy. The simple yet efficient saliency estimation technique based on regional contrast developed quickly filter out informative regions materials. This reduces unnecessary keypoint descriptors increases robustness accuracy object recognition. Our experimental results show that whole recognition process speed 19 times increase 22%. Furthermore, this stage independent choice features matching model general process. it used boost performance existing systems manner.