Joint tracking and classification of non-ellipsoidal extended object using random matrix

Many practical extended objects have non-ellipsoidal extensions. Within the random-matrix framework, a non-ellipsoidal extended object (NEO) can be approximated by multiple ellipsoidal sub-objects, each described by a random matrix. NEOs of different classes have different structures determining the relationship among the subobjects. For effective classification of NEOs, this structural information should be incorporated into the NEO models in different classes for model-based classifiers. For joint tracking and classification of a NEO using a random matrix, we propose a Bayesian framework that jointly estimates the sub-object states and extensions and obtains the probability mass function of the object class. Utilizing the structural information, the kinematic states and extensions of the sub-objects of a NEO are related to the kinematic state and extension of one reference ellipsoidal object. As such, the dynamics of a NEO can be described by a single model. Furthermore, NEOs of different classes are characterized by such models. Both the derived estimator for tracking and the classifier have a simple form. Simulation results demonstrating the effectiveness of the proposed approach are given.