Particle Filter And Its Application In Maneuvering Target Tracking Algorithm Research

In recent years maneuvering target tracking has made remarkable development asincreasingly significant usage in the military, civilian and any other fields. Leaving outof account other factors, the target tracking is a simple filtering problem. According tothe particle filter algorithm, this paper is to study the principle of the algorithm and itsapplications to maneuvering target tracking, target maneuver tracking filter.First, this article details the basic principles of particle filter. Compared withKalman filter and other filter estimating techniques, it also points out the superiority ofthe particle filter for maneuvering target tracking.Second, we give a new particle filter algorithm based on Kalman estimates. At thebeginning of filtering it uses Kalman filter to estimate target state, construct importancedensity function of the particle filter by state prediction and prediction error covarianceestimated by Kalman filter, so as to calculate the weights of particle group, so as to gainbetter characterization of the target state.Furthermore, during tracking strong maneuvering targets, continuous or strongmotor will take heave impact to the particle filter re-sampling process. Concisely to saythese strong mobility or continuous motor will lead to the re-sampling that the correctcharacterization of the target state of the particle weights turn smaller, or even beabandoned, and the particle weights even of no contribution are infinite increase, ofcourse this group of particles cannot correctly estimate the target state. So we propose anew particle filter algorithm based on genetic algorithm optimization, the algorithmretains part of the particle before re-sampling to the diversity of particles so as to elevatethe filtering accuracy by13.4%.Finally, in view of the superiority of interactive multiple models for maneuveringtarget tracking, the paper attempts to the proposed improved particle filtering algorithmapplied to the multi-model filter in order to get better tracking effect. Simulation resultsshow that, the real-time character of the improved algorithm is not high, but comparedto the basic IMM algorithm it has elevated filtering accuracy by17.5%.