A Highly Efficient Real-Time Tracking Method in Augmented Reality System

A highly efficient real-time tracking algorithm is very important for augmented reality implementation by computer vision mode. The mean-shift (MS) algorithm is widely used in object tracking for augmented reality system because of its not only speed but also highly efficient and real-time features. The traditional MS algorithm only use color feature as one of the important cues in image sequences. Even though the MS tracking algorithm cannot very suitable for complex environments, such as the background with object’s similar color, sudden light changes, occlusion types and so on. The proposed algorithm use a convex kernel function in association with the motion information to improve the MS tracking algorithm for the purpose to solve the above problems. Using these new features fusion, a robust mean-shift tracking algorithm is proposed. Experimental work show the proposed mean-shift algorithm has an optimum performance in robust and realtime object tracking for the augmented reality system.