Adaptive Model Based on Voting Probabilistic Models for Image Tracking Algorithm

摘要

This paper presents an improved template match method. We adopted the dynamic template method to match object in the tracking course, and change the size of the template dynamically according to the goal change. We adopted the self-adaptation template to upgrade the tracking tactics in order to improve the goal tracking stability. We can interpret our simple voting process in terms of a probabilistic model. This is worth doing, because it will cast some light on the strengths and weaknesses of the approach. Our generative model can be made probabilistic by assuming that the patches are produced independently and at random, assuming that the object is present. We adopted the vote model to choose the best template to track object. Finally, this paper provides the test result of the object tracking. It shows proposed approach is more effective than traditional method.