Multisensor data fusion: Target tracking with a doppler radar and an Electro-Optic camera

This paper addresses the problem of multisensor data fusion for target tracking using a Doppler radar with range rate measurements and an Electro-Optic (EO) camera. We present three fusion architectures, named FA1-FA3, to perform data fusion using the above mentioned sensors. FA1 and FA2 are distributed fusion architectures employing the information matrix fusion method with dynamic feedback. In FA1, radar and camera pseudo measurements are formed that allow us to make use of a linear Kalman Filter (KF) for the radar local filter and an Extended Kalman Filter (EKF) for the EO camera local filter. In FA2, the radar and camera measurements are used directly and therefore the system comprises two EKFs. FA3 is a centralised architecture where the data fusion is performed by way of the measurement fusion method. The final contribution of this paper is a performance comparison of these sensor data fusion techniques when making use of range rate measurements. In order to evaluate the performance of the fusion architectures, Monte Carlo simulations are performed and two filter metrics are presented: an absolute metric - the root mean squared error (RMSE) and a performance metric - the average normalised estimation error squared (ANEES). The results show that the fusion architectures presented are accurate, stable and credible.