An extended target tracking model with multiple random matrices and unified kinematics

This paper presents a model for tracking of extended targets whose extent cannot be described by a simple geometric shape such as an ellipse or a rectangle. The extended target shape is represented by a number Ns of elliptic subobjects, where Ns is assumed known. Because an extended target is a rigid body, the subobject positions must necessarily be estimated as a single state with unified kinematics, and the full covariance matrix must be estimated. In addition to the position and kinematics, for each subobject the proposed model also estimates the number of measurements generated by the shape, as well as a random matrix as representation of the size and shape. A Gamma Gaussian inverse Wishart implementation is proposed, and the state prediction and update are given. A simulation study shows the merits of the model compared to extended target modeling without unified kinematics.