<title>Tracking multiple targets in cluttered environments with a probabilistic multihypothesis tracker</title>

PMHT with Application of Road Information Based on GMTI

“…Lerro and Bar-Shalom presented debiased conversion and demonstrated that converting polar coordinates to Cartesian coordinates is superior to operating raw range-bearing measurements in Extended Kalman Filter. [5,6] …”

“…Darin T. Dunham [5] presented N-of-N initialization algorithm, in which there must be N measurement points meeting the gating criteria in the first N scans. It is necessary to set N 5 for the reliability of PMHT which makes sure the algorithm will converge to the real track There are two steps to initialize target state when N 5.…”

“…The measurement centroid is computed from (4), of which the covariance matrix derives from (5). The estimated probability is updated in (6).…”

PMHT with timing uncertainty

“…The conditional probability of the missing data, P (K x , K τ |X,θ, Z x , Z τ ), can be determined using Bayes' rule as shown in (11), which suppresses the index corresponding to the iteration i for the sake of clarity.…”

“…The PMHT uses EM to model the assignment of measurements to targets as hidden variables, and estimates target states by taking the expectation over the assignments. The PMHT has evolved since it was introduced, with many extensions including a track management system to do track initiation, termination, and merging [9,10]; an extension to deal with cluttered measurements [11]; and the Turbo PMHT, which improved the performance of the PMHT substantially using turbo decoding [12]. Many variants of the PMHT algorithm-including homothetic PMHT, manoeuvre-based PMHT, and various PMHTs with different measurement/target association models-are compared in [13].…”