Research Of An Improved Particle Filter Algorithm And Its Application In Target Tracking

Because of its great potential in the non-linear and non-Gaussian areas, particle filter has received attention widely in the filed of nonlinear filter. Comparing to other traditional filters, particle filter is easy to use and implement, so it receives extensive application in many reserch fileds.Particle filter is a practical method in resolving bayes possibility and adapt to any nonlinear system, it implements recurrence bayes estimation by using nonparametric Monte Carlo method. But traditional particle filter has a few of shortcomings, such as typical particle degeneracy, and particle diversity caused by re-sample. So far, although there are many improved programs, they do not resolve the problems perfectly. Therefore, it has important significance to enhance the efficiency of particle filterThe main contents and novel parts of the thesis are as follows:1. The paper incorporates adaptive genetic algorithm into particle filter to overcome the drawback of the filter, using genetic operators, such as crossover, mutation and selection to operate particle, until the particles implement better than before, the diversity of particle is enhanced and the search region of particles is enlarged.2. In order to meet its application requirements in particle filter, adaptive genetic algorithm will be experienced a series of improvements. In particle filter, algorithm aims to obtain an optimal population not the best individual as traditional adaptive genetic algorithms. Therefore, we need to adjust the variable adjustment strategies in the traditional adaptive genetic algorithms according to the prior knowledge of each particle.3. In this paper, we applicate our improved particle filter in the field of computer vision. At first we achieved the tracking algorithm by standard particle filter and by Improved Particle Filter. Then, according to the experimental results, our novel particle filtering is better than standard one.