Natural feature tracking for augmented reality

Abstract

A robust framework for 6DOF camera pose tracking based on natural features in anarbitrary scene was developed for online Augmented Reality applications. Crucially, the method is based on calculating camera pose relative to pre-captured reference images. This prevents a gradual increase in the camera position error. Point features in the current frame are first matched to two spatially separated reference images. This wide baseline correspondence problem is overcome by matching to the previous image frame and transferring point positions to thereference images. Deviations from the two-view and three-view constraints between the reference images and current frame are then minimized as a function of the camera motion parameters. This calculation is stabilized using a recursive form of temporal regularization similar in spirit to the Kalman filter. Using this system, camera pose is reliably tracked over hundreds of image frames and virtual objects are realistically integrated with only slight jitter.