Fast unscented Kalman filter for projective tracking
Bensoukehal Ali, Mokhtar Keche, Abdelkrim Meche · International journal of imaging and robotics · 2019
This paper presents a novel approach to solve the problem of estimating and tracking a moving target, observed from projective viewpoint. A real-time tracking is achieved by a newly designed filter based on the steady state linear Kalman filter, known as αβ filter. The nonlinearity problem introduced by the projective nature of the observed data is handled using unscented transform, instead of the standard first order approximation. The effectiveness of the proposed filter is demonstrated and compared with that of existing methods by Monte-Carlo simulations.