Research On Target Tracking Algorithm Based On Adaptive Kalman Particle Filter

Particle filter is an effective filtering technique developed rapidly in recent years. It has been widely used in many fields such as image processing and target tracking because of its great advantage over solving the problems of nonlinear dynamic system. It can be used as a new method to realize the recursive Bayesian filtering algorithm with the aid of Nonparametric Monte Carlo simulation. New theory on particle filter has developed quickly and a lot of effective algorithms appeared. Nowadays, particle filter has become a research hotspot in the area of filtering theory worldwide.Firstly, the thesis introduces the Extended Kalman Filter (EKF), Unscented Kalman Filter (UKF), Genetic algorithm and Particle Filter algorithm. The algorithms mentioned above have good abilities in target tracking, but they all suffer from deficiency of imprecision and maladjustment in the environment of high mobility, non-Gauss and low signal-to-noise ratio.Secondly, an improved adaptive UKF algorithm is presented. The UKF algorithm is sensitive to the initial value and the state model deviation. To solve the mismatch between the carrier's movement and the filtering model, a fast adaptive algorithm is adopted in the modified UKF algorithm.Thirdly, comparison is made between the unscented Kalman particle filter algorithm based on adaptive algorithm and the conventional method. Experiments have shown that the former can overcome the weakness of traditional particle filter algorithm by using the state transfer probability as particle filter proposal distribution. The proposal distribution generated by adaptive UKF particle filter and sampling particle wherein make the probability of particle distribution closer to the state probability distribution. Many other techniques are also adopted, such as the GA-MCMC algorithm in the process of resampling particle filter, global searching by genetic algorithm, the convergence of the Markov Monte Carlo method, and so on. Simulation results demonstrate that the adaptive UKF algorithm based on the GA-MCMC can greatly reduce the particle degeneracy and impoverishment in the iterative process.