The Study On Algorithm Of Object Tracking Based On Color Feature

Along with the development of computer vision technology energetically, object tracking technology in application and requirement both has a further promotion. Tracking method based on detection has been difficult to meet demand of engineering practice.Based on the connection between the image sequences, now object recognition and correlation method has been more widely adopted. And the study of identification method of tracking has many achievements. However, practical experience has shown that object tracking technology is currently far from mature ,and there are still many problems to solve in order to build a robust and real-time tracking system.Mean-shift is a no parameter density estimation algorithm,with good convergence and fast speed when tracking target.But due to the nature of single-peak search algorithm,the target tracking system is not robust in many cases.The particle filter is to completely describe the posteriori probability distribution by a group of weighted particles.And its multimodal search property makes it Superior to the Mean-shift algorithm in the tracking process.But the computational complexity results to poor realtime apply.Based on the research of basic theory and knowledge in object tracking ,this paper chooses color information for target features, and implements the Mean-shift and particle filter object tracking algorithm in MATLAB respectively.Finally, the two algorithms are combined, using the method of convergence particle in Mean-shift other than the traditional resample method of particle filter, in the process of particles and the convergence of using adaptive method of particle area. Experiment results show that this algorithm can reduce the particle numbers and improve tracking speed.