The Research Of Particle Filter Algorithm In Target Tracking

Target tracking which has an extremely wide application both in military and civilian areas has been a hot topic of scientific research. There are two key elements in target tracking: the establishment of accurate tracking model and the design of precise tracking algorithm. Under the framework of Bayesian filter, particle filter algorithm utilizes all the measurements of system to recursively estimate the posterior probability distribution of state variables in order to get the best state estimation by converting target tracking problems into Bayesian estimation problems. Compared with traditional filtering algorithms, particle filter can deal with the state estimation problem more effectively without the limit of linear system, system dimension or noise distribution. Therefore, particle filter algorithm has been widely used in the field of target tracking. With the development of science and technology, the maneuverability of moving target is getting stronger. However, the interactive multiple model which has the advantages of structuration, fine stability and high tracking accuracy can describe target state accurately. Thus,the combination of particle filter and interactive multiple model can improve the tracking accuracy in maneuvering target tracking.The main work in this dissertation is listed as follows. First, according to the sufficient research of particle filter and traditional resampling algorithms, this dissertation puts forward an improved particle filter algorithm based on resampling. The algorithm adds the pre-treatment and weight linear optimal combination operation before resampling in the filtering process. Pre-treatment can move the small weight particles to high likelihood zone by using the mean of particle pair to replace small weight particles while increasing its weight. This process can reduce the variance of weights between particles, increase the number of effective particles, decrease the times of resampling, improve the performance of state estimation and guarantee the diversity of particles to some extent. Utilizing weight linear optimal combination before resampling can increase the number of replicated particles,alleviate the problem diversity loss of particle and improve the performance of state estimation. In conclusion, the improved particle filter has a better performance of state estimation. The simulation results of MATLAB show the effectiveness of the algorithm. Second, this dissertation proposes the combination of improved particle filter and interactive multiple model based on the detailed study of interactive multiple model. And the combined algorithm is applied in maneuvering target tracking. The validity of the algorithm is verified by two simulation scenarios.