Research Of3D Content Generating System Based On Virtual-real Fusion

3D display technology is promoting the revolution of the image and video industry, it is also affecting all aspects of our life. Along with the development of3D display technology,3D content production technology becomes a new hot spot. People are looking for3D content production methods which are convenient and cheap. Among them, the method based on fusion reality is an important research direction.The3D technology and virtual-real fusion technology brings great changes to people’s perception approach, the combination of them represents the general trend. On the one hand,3D technology provides a more realistic display mode for the virtual-real fusion content. On the other hand, virtual-real fusion technology brings a new method to generate3D content. The3D content generating system based on virtual-real fusion technology has broad market prospect and great research value.The purpose of this paper is to study key technologies in the3D content production system based on virtual-real fusion. Firstly, the camera tracking algorithm for indoor scenes is studied. Secondly, it is discussed that how to achieve reliable virtual viewpoint contents using the images captured by a monocular camera. Lastly, it is researched that how to obtain realistic3D virtual-real fusion contents by mixing virtual information and real scenes. The main achievements of this paper include:(1) For indoor scenes, a camera tracking algorithm is implemented based on the reconstruction of scenes, and stable real-time tracking is obtained. Especially for motion interference scenes, a moving object filtering algorithm is presented to improve the tracking robustness.(2) A virtual viewpoint rendering algorithm based on scene models is presented. Scene reconstructed models is utilized to solve problems in traditional depth image based rendering algorithms and high quality rendering is attained.(3) Two viewpoints virtual-real fusion is achieved. Stable3D registration effect of virtual objects is obtained due to the stable tracking performance. A simple and effective method of occlusion judgment and collision detection is put forward at the same time, achieving realistic3D format fusion contents.