WEAR++: 3D model driven camera tracking on board the International Space Station

摘要

We present WEAR++, a wearable augmented reality system consisting of a head mounted display, a camera and an inertial measurement unit. This paper focuses on the visual camera tracking system developed for WEAR++. Using a 3D model of the scene, we first create a map of 3D-2D correspondences in an off line mapping procedure. During on line operation, we match features from a new image to the database, and track the camera pose with an Extended Kalman Filter using the recovered 3D-2D correspondences. By using robust local features (SURF) and a frustum culling algorithm, we demonstrate that we are able to track the pose even for jerky motions and blurry images. Furthermore, we explain how the system was utilised by astronaut Frank De Winne on board the International Space Station for performing maintenance tasks.