Improvement and Demonstration of Image Sequence Tracking Algorithm Based on Particle Filter

摘要

Regarding the problem of tracking the moving targets in image sequences under complex background, this paper proposes an improved particle filter image tracking algorithm. Using the results of genetic algorithm, this algorithm has achieved the resampling of particles by means of selection, crossover and mutation and so on and solved the problems of degradation and deprivation of particles faced by the particle filter. Because the algorithm uses the global optimization feature of the genetic algorithm, it has a strong robustness. Meanwhile the particle filter can implement the optimal estimation of the nonlinear, non-Gaussian state space model When applied in target tracking, the particle filter has better over occlusion ability. Results show that this algorithm presents good performance of the state estimation and can well achieve the target tracking in very complex image sequences.