Ellipse Fitting Based Approach for Extended Object Tracking

With the increase of sensors' resolution, traditional object tracking technology, which ignores object's physical extension, gradually becomes inappropriate. Extended object tracking (EOT) technology is able to obtain more information about the object through jointly estimating both centroid's dynamic state and physical extension of the object. Random matrix based approach is a promising method for EOT. It uses ellipse/ellipsoid to describe the physical extension of the object. In order to reduce the physical extension estimation error when object maneuvers, the relationship between ellipse/ellipsoid and symmetrical positive definite matrix is analyzed at first. On this basis, ellipse/ellipsoid fitting based approach (EFA) for EOT is proposed based on the measurement model and centroid's dynamic model of random matrix based EOT approach. Simulation results show that EFA is effective. The physical extension estimation error of EFA is lower than those of random matrix based approaches when object maneuvers. Besides, the estimation error of centroid's dynamic state of EFA is also lower.