Adaptive Measurement Noise Covariance Matrix R for JPDAF based Multitarget Tracking

Abstract

In multitarget tracking, several targets of interest are being tracked simultaneously with the help of any optimal estimator. Kalman Filter (KF) and Extended Kalman Filter (EKF) have proved to be very good estimators. Multitarget tracking finds its applications in diverse fields like pattern recognition, computer vision, radar tracking, robotics, etc. Several algorithms have been implemented for multitarget tracking including Probabilistic Data Association Filter (PDAF), Joint Probabilistic Data Association Filter (JPDAF), Nearest Neighbor Standard Filter (NNSF), Global Nearest Neighbor (GNN), Neural Networks (NNs), etc. Joint Probabilistic Data Association Filter (JPDAF) is the multitarget version of Probabilistic Data Association Filter (PDAF), in which joint association probabilities are computed and tracks are then updated based upon theses probabilities. Measurement noise covariance matrix R in Kalman filter needs to be transformed from polar to cartesian coordinate system. The optimal value of R should be calculated for the good performance of filter. In this paper, measurement noise covariance matrix R has been computed using transformation and more than 80% of the desired results have been achieved by performing tracking using JPDAF algorithm.