Random Matrix Based Extended Target Tracking With Orientation: A New Model and Inference

In this study, we propose a novel extended target tracking algorithm which is capable of representing the extent dynamic objects as an ellipsoid with time-varying orientation angle. A diagonal positive semi-definite matrix defined to model objects' within random framework where elements have inverse-Gamma priors. The resulting measurement equation non-linear in state variables, and it not possible find closed-form analytical expression for true posterior because absence conjugacy. We use variational Bayes technique perform approximate inference, Kullback-Leibler divergence between minimized by performing fixed-point iterations. update equations are easy implement, can be used real-time applications. illustrate performance method simulations experiments real data. proposed outperforms state-of-the-art methods when compared respect accuracy robustness.