Research On Object Tracking Via The Collaboration Of Discriminative And Generative Models

In recent years,object tracking has become an active research topic because it's one of the most challenging topics in computer vision.In order to improve the effect of the object tracking and take the advantages of the discriminative and generative models,we proposed the collaboration of the discriminative and generative models for object tracking.Then we want to relieve the problem in the object tracking process,these problems are occlusion,motion blur,fast motion etc.The research work of this paper includes two aspects.The one is the object tracking algorithm based on the Spatio-Temporal Context algorithm and the Mean Shift algorithm.The other one is the object tracking algorithm based on Kernelized Correlation Filters algorithm and the Mean Shift algorithm.The main work and innovations are as follows:(1)When the object undergoes heavy occlusion,the Spatio-Temporal Context(STC)algorithm can track the object accurately,but the Mean Shift(MS)algorithm is shaking in this situation.After occlusion,the Mean Shift algorithm can track the object again.However,the Spatio-Temporal Context method cannot finish it.In order to make full use of advantages of the two algorithms,we develop a new algorithm STCMS to combine these two algorithms.Frankly,our algorithm can solve the problem of heavy occlusion.We present thorough experimental results on a number of challenging sequences in terms of efficiency,accuracy and robustness.(2)In order to solve the problems of motion blur and fast motion,a new robust object tracking algorithm KCFMS using the Kernelized Correlation Filters(KCF)and the Mean Shift algorithm is presented in this paper.In every frame we do these work: First,we give the initial position and size of the object and use the Mean Shift algorithm to obtain the position of the object.Second,the Kernelized Correlation Filtering algorithm is used to obtain the position of the object in the same frame.Third,we use the Cross Update Strategy to update the object models.In order to improve the tracking speed as much as possible,our object tracking algorithm is only one layer.This hybrid algorithm has a good tracking effect on the target fast motion and motion blur.We present thorough experimental results on a number of challenging sequences in terms of efficiency,accuracy and robustness.