Tracking Registration Algorithm for Augmented Reality based on Random Fern

For markerless tracking registration methods, there exist a lot of problems like tracking registration failure and slow speed in the complex environment. To solve the problem, a random fern classifier is applied to the process of the markerless augmented reality tracking registration. This method uses a target object image in a real scene as a template, and the target of each frame of image is detected by the random fern classifier. By estimating the three-dimensional pose, the registered virtual object can be rendered to implement the augmented reality tracking registration. This method can solves the problems of tracking registration failures due to the change of ambient lighting or the targets were occluded. It has better real-time performance than the traditional wide baseline matching algorithm.