Multiple Extended Target Tracking Based On Gaussian Mixture Probability Hypothesis Density Filter

With the increasing development of sensor technology, the technique of extended target tracking plays a more and more important role in the military and civilian fields like the missile targeting, battlefield surveillance, intelligent transportation and medical diagnostics.For the extended target tracking problems under complex environment, since a single extended target at each sampling period potentially gives rise to more than one measurement,measurement partition is a matter of prime problem to firstly resolve. Whether measurement partition is correctly divided directly affect the subsequent targets state estimation. Therefore,the study of a more correct and efficient measurement partition method is of great theoretical significance and application value. Based on the extended target measurement partition, this dissertation mainly involves researches about the achievement of the extended target tracking using the Gaussian mixture probability hypothesis density filter algorithm.For the measurement partition problems of extended targets, this paper uses two methods for dividing the measurement set : Distance partitioning and K-means++ partitioning. Firstly,the two measurement partition method divided the extended target set. Then In Matlab environment, Using the Gaussian mixture probability hypothesis density filter to simulate extended target tracking under the two measurement partition methods, simulation results show that compared with the K-means++ partitioning method, distance partitioning performance better in extended target tracking.In complex environment, since multiple extended targets generate huge number of measurements in a sampling instant and the existence of clutter, the observation has uncertainty. It makes the measurement partition process of the methods mentioned above can’t get ideal results. In order to improve the extended target tracking accuracy by using the measurement set partition. So this dissertation uses a new extended target measurement partition method base on the Affinity propagation clustering. The Affinity propagation clustering measurement partition method firstly separates the clutter measurements from target measurements by density analysis techniques, and then use affinity propagation clustering technology to partition the extended target measurement, thus, reducing the number of partition. Finally, Based on Gaussian mixture probability hypothesis density filter algorithm, it can achieve the extended target tracking by using the algorithm. simulation results show that, compared with the first two measurement partition, the algorithm enables extended target tracking performance has greatly improved.