Hybrid Tracking for Augmented Reality

In augmented reality, the real world is augmented by virtual, computer generated objects. If real and virtual objects are not aligned properly, the experience of the user will not be satisfying, as the illusion of a common world consisting of both real and virtual objects will not be present. To achieve this precise alignment, it is necessary to track the user’s position and orientation with high accuracy. In addition, real-time operation of such a tracking system is a necessity, otherwise precise alignment while the user is moving around is not possible. In this thesis, we present a hybrid tracking system that combines a standard magnetic tracker and an optical tracking system. This combined system is more precise than a standalone magnetic system and both faster and more reliable than a standalone optical tracking system. In order to satisfy the need for real-time operation, data from the magnetic tracker is used to predict feature locations for the optical tracking system. Feature positions are then determined using a new, fast tracking primitive which tracks lines with sub-pixel accuracy. It is based upon the XVision tracking library and uses only a small search area within the image. From these feature locations in the image, position and orientation of the camera, and thus, the user are computed. This hybrid tracking system is capable of tracking approximately twelve features at a rate of ten frames per second. In addition to the tracking system we discuss the calibration requirements in order to achieve a satisfying augmentation. Calibration comprises camera calibration, where both interior and exterior orientation are determined, room calibration, where transformations among the various coordinate systems are established, and helmet calibration, where the head mounted display is calibrated with respect to the user. In the experimental section of this thesis we compare the new sub-pixel edge tracking primitive with the original XVision primitive. We give results on the accuracy of the optical tracking system and contrast these with data