Augmented Reality Tracking Prediction Method Study

Augmented Reality (AR) is a process overlaying the virtual scene generated by the computer on real-world scene. AR technology is widely applied in many areas, such as military, medical, education, culture and so on. With the development of Augmented Reality technology, the real time performance of tracking and registering is higher than before. In order to improve the efficiency of tracking and registering, studying tracking prediction methods of Augmented Reality is very important.Due to heavy calculation of tracking in Augmented Reality real time performance of registering of virtual objects is often effected. For solving this problem, tracking method on the basis of goal prediction is presented in this paper. In this method, according to motion correlation among continual frames camera's pose is estimated by reducing the area of searching. Main work in this paper follows as:1. Key technologies of Augmented Reality are studied. Tracking and registering process of these common algorithms are researched.2. Algorithms of motion estimation are studied. Different algorithms of motion estimation are analyzed. Various matching criterion functions, search strategies and search efficiency of block matching motion estimation are discussed. And then advantages and disadvantages of these block matching motion estimations are analyzed.3. For four Step Search method and Diamond Search being easy to fall into the most superior of part, an algorithm of motion estimation on the basis of self-adaptive template is proposed in this paper. By setting searching window, wide range template or the small one is determined according to the value of criterion function. During this process the best matching point is detected and based on this the distance and direction of camera could be estimated.4. For verifying the algorithm it is applied into Augmented Reality system. Usually, real time performance of tracking and registering is influenced by heavy tracking calculation of camrea's pose. Through using the proposed algorithm, the moving distance and direction of camera can be estimated. Experiments show that this algorithm helps improve the real time performance of tracking and registering.