Counting targets with MHT and CPHD

This paper studies the counting-targets limit of the multiple-hypothesis tracking (MHT) and cardinalized probability hypothesis density (CPHD) solutions to the multi-target tracking (MTT) problem. The solutions are compared with a direct Kalman filtering (KF) solution to the counting-targets MTT problem, whereby we assume a continuous state (number of objects) and assume linear Gaussian measurements. While the enhanced MHT solution - the cardinalized MHT (CMHT) - performs well, it does not match the performance of the KF and of the CPHD. In future work, we will assess these solutions with an RMSE performance metric and examine whether there is any sub-optimality in the CPHD solution to this problem.