Adaptive Target Tracking Algorithms Based on Particle Filter

Particle filter algorithm is a kind of algorithm which leverages the Monte Carlo simulation to complete a Bayesian recursive process. In this paper, we illustrate the principle of particle filter based on nonlinear and non-Gaussian system's state filtering. Mean shift [8] algorithm is a method based on the optimal gradient descent, which searches the target by means of iteration, realizing the moving target tracking. However, Mean Shift method cannot deal with complex backgrounds and targets under occlusions. First, we improve image histogram, and then propose a target model based on statistical histogram distribution, and finally combined these two methods together effectively by this model. According to the tracking process, adjusting the parameters adaptively can process the effects brought by light changes or shelters in the image sequence relatively well. The result of the simulation shows that the particle filtering algorithm is an effective solution to the problem about how to track a target whose movement is non-Gaussian.