In this paper we present a novel visual-inertial 6-DOF localization approach that can be directly integrated in a wearable immersive system for simulation and training. In this context, while CAVE environments typically require complex and expensive set-up, our approach relies on visual and inertial information provided by commodity hardware, i.e. a consumer monocular camera and an Inertial Measurement Unit (IMU). We propose a novel robust pipeline based on state-of-the-art imagebased localization and sensor fusion approaches. A loosely-coupled sensor fusion approach, which makes use of robust orientation information from the IMU, is employed to cope with failures in visual tracking (e.g. due to camera fast motion) in order to limit motion jitters. Fast and smooth re-localization is also provided to track position following visual tracking outage and guarantee continued operation. The 6-DOF information is then used to render consistently VR contents on a stereoscopic HMD. The proposed system, demonstrated in the context of Construction, runs at 30 fps on a standard PC and requires a very limited set-up for its intended application.
|Title of host publication||VRST '14 Proceedings of the 20th ACM Symposium on Virtual Reality Software and Technology|
|Place of Publication||New York|
|Number of pages||4|
|Publication status||Published - Nov 2014|