Algorithm Research On Target Tracking

In many fields, like in the field of the battlefield surveillance, antiaircraft system in a war, and the civil traffic control, machine intelligence, medical equipment, target tracking is a basic or an important problem. As its application in more fields, many new techniques are introduced to the target tracking for more complicated situation. While, one core part in target tracking is filtering algorithm. Here, the algorithms of Kalman filter, extend Kalman filter (EKF), particle filter, and the interacting multiple model (IMM) based these filters are studied.The principle of target tracking is introduced first in this paper, analyzing the classical target tracking method based on Kalman filter. Through a simulation, it is known that target tracking method based EKF performs well when tracking a target in a linear system or with small maneuver but bad when in a strongly nonlinear system or with high maneuver. Also, the Jacobians matrix of EKF is very hard to calculate in tracking system with multidimensional state. To solve these problems, the particle filter, a very popular algorithm in recent years, is introduced. It is found that the particle filter indeed performs superior to EKF when tracking a target in a strongly nonlinear system or with high maneuver. However, the algorithm of particle filter has an expensive computation because of its principle. Finally, the algorithm of IMM for the maneuvering target is studied, and the algorithms of Multi-rate IMM Particle Filter (MRIMMPF) and IMM Kalman-Particle Filter (IMMK-PF) are proposed. The MRIMMPF, to which the multi-rate technique is introduced based on the IMMPF algorithm, is hoped to keep the performance of the IMMPF with less computation. The IMMK-PF, combining the advantages of Kalman filter and particle filter, is expected to improve the algorithm's robustness and computational efficiency. The comparison experiment with IMM based on EKF, IMMPF demonstrates the validity of the two algorithms.