Joint tracking and classification of extended object using random matrix

Most practical extended objects can be classified by their size and shape. The random-matrix approach to extended object tracking provides efficient estimation of both the centroid state and the extension. For effective classification of objects, however, prior size and shape information of the objects needs to be sufficiently modeled into the random-matrix-based framework. For joint tracking and classification of an extended object using a random matrix, we propose a Bayesian framework within which the probability density function of the object state and extension and the probability mass function of the object class are obtained jointly. Only measurements of scattering centers are needed in this framework. The size and shape properties distinguishing objects of different classes are treated as constraints and integrated into the framework as pseudo-measurements. Online orientations of the objects are obtained by a maximum likelihood method. Both the derived estimator and the likelihood for classification have a simple closed form. Simulation results demonstrated the effectiveness of the proposed approach.