Particle Filter Algorithm And Its Application In Target Tracking

Kalman Filter and Extended Kalman Filter are the most typical filter algorithms in target tracking domain, of which the former is adaptive to linear and the latter is used in nonlinear system. Kalman Filter is an optimal filter algorithm in the Minimum-Mean-Square-Error sense, While the non-Linearity of the system is not extreme strong, EKF can achieve approximately optimal filter effect. Kalman Filter and Extended Kalman filtering performance will descend or even diverge when non-Gaussian distribution occurs.The paper focus on nonlinear problems, describes what is the particle filter. Particle filter attract more reserchors in kinds of field. Particle filter method can deal with the non-linear and non-Gaussian noise system, it is used in communications, radar tracking, image treating, computer vision, fault detection and so on.Particle degradation is inevitable phenomenon in the particle filter algorithm, the solution is mainly choosing the proper proposal distribution and Resample. There is no a common choice of proposal distribution method, each one has its advantages and disadvantages. This paper describes an improved particle filter algorithm. The algorithm uses unscented kalman filter to generate particles for estimating the state, and adds fading memory factor to the update process, in this way it can weaken the dependence of the filter on the history information to enhance the current measurement correction function to filter information. Thereby it generate a preferred proposal distribution function, and then the preferably particle degradation is suppressed. The article through the matlab simulation compares the improved filter with other particle filter algorithm, the experimental results show that this algorithm is a kind of effective algorithm, and it can achieve better estimated.Multiple maneuvering target tracking technology in military, civil and other fields have a wide range of applications. Data connection link is much more important link in maneuvering target tracking problem, namely judges a specific measurements belong to which a goal, and to allocate all or part of the new measurement data to established track.This paper combines the nearest-neighbor method and the particle filter algorithm applied to target tracking system, and compares particle filtering with the improved filtering algorithm for multi-tracking performance, it is can be seen from the simulation diagram that the improved filtering algorithm can match the real trajectory, and has high estimation precision.