Real-time surface mapping and tracking is the core technology behind augmented reality applications. The growing trend of integrating depth sensors into mobile devices provides a great opportunity for enhancing the experience of interaction with reality. This work presents a system that utilizes the RGB-D cameras of mobile devices, for tracking and reconstructing dense surface in real-time. This task is challenging because the low computing power and memory capacity on consumer mobile platform limit the scale and precision of the reconstruction. To scale up scene size and provide smooth integration experience, we implemented memory efficient data representation and caching methods, and also make the integration and visualization processes highly parallel. This thesis illustrates...[ Read more ]

Real-time surface mapping and tracking is the core technology behind augmented reality applications. The growing trend of integrating depth sensors into mobile devices provides a great opportunity for enhancing the experience of interaction with reality. This work presents a system that utilizes the RGB-D cameras of mobile devices, for tracking and reconstructing dense surface in real-time. This task is challenging because the low computing power and memory capacity on consumer mobile platform limit the scale and precision of the reconstruction. To scale up scene size and provide smooth integration experience, we implemented memory efficient data representation and caching methods, and also make the integration and visualization processes highly parallel. This thesis illustrates the implementation of each pipeline stage, including the underlying data structure, pre-processing, tracking, integration, rendering and caching. The overall performance evaluation and its application are also demonstrated with a conclusion.