Tracking-Verfahren

Tracking procedure

In order to realize an optimal integration of the virtual contents into the environment for visual augmentation, a correct geometric and photometric registration of these objects is required. For the exact determination of the position and attitude of the observer or his augmented reality system, we explore novel feature-based computer vision methods. These methods are based solely on webcams or smartphone cameras in order to determine the most exact position in real time.

Among other things, we use distinctive features that are visible in the real environment. These features have to meet a wide range of requirements so that they remain detectable from different positions and under varying lighting conditions. Thus, our department is investigating, among other things, the well-known SURFund FAST detectors, each of which detects different features with specific characteristics (Figure 2). The exact position and location in space can be determined by the correspondence of points from the 3D environment and points in the 2D camera image (Figure 3). Feature-based computer vision methods are particularly computationally intensive and therefore require a high degree of optimization to be able to use them on mobile devices such as smartphones.