| Summary: | 碩士 === 國立臺灣大學 === 資訊工程學研究所 === 87 === To solve the registration problem of virtual object and real image is a main research issue in augmented reality (AR). The purpose of AR is to allow the users to feel that they can not distinguish the differences between virtual objects and real images. In this thesis we adopt the method based on “parameterized cuboid structure (PCS)” for the estimation of the camera parameters and for the composition of the virtual objects and real image. In our approach, the camera needs not to be calibrated in advance. We assume a simple camera model, and a user need to select a PCS in the image for the estimation of the parameters of the camera. Our method can also provide a useful functionality — the auxiliary line, which can help the user to select a reasonable PCS in the image in an easier way.
Based on the PCS method, we developed an AR system which can not only integrate virtual objects into a single image but can also integrate virtual objects into image sequences. The main feature of our method is that if the user can select a PCS in the scene (explicitly or implicitly), our system can then integrate virtual objects into the image. In addition, the user can interactively place virtual objects (either static or moving) in the image with respect to the coordinate system of the PCS.
We can also use this method to integrate virtual objects into a panoramic image. In fact, this problem is simpler because the intrinsic camera calibrations are recovered as part of the panorama construction. In our system, the user needs only to select four points in the cuboid structure, the extrinsic camera parameters can then be estimated automatically. After the extrinsic camera parameters have been estimated, the virtual objects can then be successfully integrated into the panorama.
|
|---|
