Visual Tracking With Improved Quantum-Behaved Particle Swarm Optimization

Visual tracking is one of the most important applications in computer vision. It is widely used in intelligent transportation, virtual reality, automated surveillance, human-computer interaction and so on. Meanwhile, due to the complexity of the scene where the object lies in, it is still a research problem to design robust, real-time tracking method. Therefore, visual tracking is of great value in research and has huge space for development.Since the tracking process can be formed as a dynamic optimization problem, optimization algorithms used in optimization problem are widely used in visual tracking. The particle swarm optimization(PSO) algorithm is a hot topic in Swarm Intelligence recently, and can be used to solve optimization problem. So the PSO based tracking method can be a new research field in visual tracking. Some researches have shown effectiveness of the usage of PSO in visual tracking.The particle swarm optimization(PSO) based tracking method shows good performance. However, PSO has been proved to be a local optimal algorithm, i.e. it is easy to converge to the local optimal solution. Based on the quantum theory and PSO, quantum-behaved particle swarm optimization(QPSO) can overvome the shortcomings of PSO while maintaining the advantages of PSO. In this paper, we try to introduce a tracker based on QPSO. However, QPSO also has the disadvantage of losing diversity, which may cause damage in tracking. Inspired by Differential Evolution(DE) algorithm, we present a new QPSO algorithm named DEQPSO that incorporates mutation, crossover and selection operation. At the same time, we design a patch-based and adaptive observation model based on Gaussian Mixture Model. Based on the researches above, a new tracking method based on DE-QPSO is introduced. We conduct numerous experiments, and the results have shown the effectiveness of our method.