An improved Particle filter tracking algorithm

In target tracking, if the dynamic model satisfies the Kalman filter assumptions, Kalman filter is optimal, Particle filter is a second-best. Usually, systems are often unable to meet the best, at this time particle filter is usually better than any other filtering method. In order to solve the degradation and deprivation of particle filter in Iteration. This article introduces crossover and mutation operations in the process of sampling and resampling. As the Gabor wavelet is not sensitive to the geometric distortion, brightness change, and noise in the process of describing the objectives, and it is able to achieve a stable tracking for the target with Partial occlusion. So this article construct Gabor wavelet feature template, proposed an improved Particle Filter Algorithm, and implement stable tracking to the target in different contexts.