Research On Particle Filter Algorithm And Its Application In Target Tracking

As a non-linear and non-Gaussian filtering method,particle filter(PF)can be applied in different fields.The particle filter algorithm has a good robustness.In the target tracking process,accuracy and real-time performance are basic requirements.As a filter technology for nonlinear and non-Gaussian systems,particle filter can perform better filtering and track nonlinear moving targets,and it has a good robust.The particle filter algorithm can track the targets in a complex environment,but it can't add the actual observation information to the real-time tracking,so it causes a lot of computational waste,and it can't effectively express the posterior probability density function.Resampling can solve the problem of particle degradation,but it causes the diversity of particles reduce.When the tracking accuracy is not enough,the system may not converge.The differential evolution algorithm is an evolutionary algorithm.It optimizes the offspring through continuous iterative updating,the differential evolutionary is feasibility to solve particle degradation,but the differential evolution mutation rate is fixed,after several iterations,the differences between offspring samples become smaller,the diversity of sample is reduced.A mutation-adaptive differential evolution algorithm is introduced.In the mutation process,the adaptive mutation rate makes the large-weight particles changes little while the small-weight particles changes large,so the small-weight particles can have an optimal state.The particle samples are optimized and the particles are diversity,the distribution of particle sets approaches the optimized posterior distribution.The mutation-adaptive differential evolution process optimizes the particle distribution,increases the accuracy of the estimation.It can achieve the same accuracy as a traditional particle filter algorithm with fewer particles and the resampling is replaced with a differential evolution process,which can solve particles filtering requires a large number of particles in the tracking process and eliminates the need for resampling.The differential evolution algorithm of adaptive mutation rate improves the particle degeneration problem in the particle filter algorithm,it can be applied in the different motion models.The experimental simulation results show that the differential evolution of adaptive mutation particle filter algorithm has higher accuracy than the particle filter algorithm improved by differential evolution algorithm.Introducing the differential evolution of adaptive mutation rate can alleviate the problem of particle degradation existing in the algorithm.The accuracy of the algorithm is improved,and the resampling process is replaced by a differential evolution process,which can achieve higher accuracy with fewer particles.The algorithm has a shorter running time and better real-time performance.