Research On Transforming State Space Extended Kalman Filter And Application In Track Fusion Problem

Track fusion is a very important aspect of the multi-sensor data fusion. It includes target tracking, data association and fusion of related track. Association and fusion are both based on target tracking, so it is of great significance to do research on target filter-tracking algorithm.We introduce the basic theory of multi-sensor data integration firstly, expound development of Kalman Filter (KF for short) theory and the basic equations of KF, we expound deducing-process of Extended Kalman Filter (EKF for short)equations in this paper. Then we lucubrate Transforming State Space Extended Kalman Filter (TSS-EKF for short). Improved TSS-EKF algorithm is proposed in this paper, at the same time we deal the problem of track fusion with this improved algorithm. Three mathematical models about track filter-tracking, data association and integration of related tracks are established. We carry on corresponding simulation. From the result of simulation, it is clear that we get smoother curve of track and smaller error, we gain exact test of track correlation, we find satisfying result of data fusion.New studies of theory and application are as follows:(1) The definition and of new state space function is putted forward in this paper. We give the construction process of this new function and put forward improved TSS-EKF algorithm based on the idea of transforming state space. Through analysis of an example, we point out the limitation of old TSS-EKF algorithm, so the advantage of improved TSS-EKF algorithm is clear.(2) The corresponding new state space function is established facing to real track data which has strong flexibility. We carry real time simulation of track filter-tracking by using improved TSS-EKF algorithm, then we get smoother curve of track. (3) As the error of track data is rather big, we study if tracks are correlated or not by definition vector quotient of two tracks state.Then a new association test model was put forward by using hypothesis testing. It is a new application of hypothesis testing when we test track association.(4) Integration formula of arbitrary relevant tracks is extrapolated, and mathematical models of data fusion are established. We carry on corresponding simulation of correlated track.