A Study on the Active Vector Model for Object Tracking
In this thesis we propose a new active vector model which is able to track objects with an algorithm. First of all we classified a few basic shapes as the modes of the tracking object, which were learned by the principle components analysis, and then we extracted the representative feature vector and the minimum shape parameters. And we reorganize the sequence of basic shape change to the shape change based on the feature point vector. We modeled the object of target tracking and its moving using both the feature position vector and shape change vector obtained by the above process. The proposed method generates parameterized values based on the moving pattern of the object, provides better stability of the local structure than other models, and decreases the cost of convergence duration, which is the weakness of model-based tracking algorithms.