A Study on Multiscan Automatic Target Tracking Algorithms in Cluttered Environment Using a High Pulse Repetition Frequency Radar

Abstract

The target tracking is a hybrid research field that integrates both technologies from the state estimation and the data association to account for the measurement origin uncertainty in addition to their inaccuracy. As one of the typical target tracking applications, the high pulse repetition frequency (HPRF) Doppler radar tracking is pervasively used in the airborne surveillance system. However, there exist two main challenges for the HPRF radar tracking in cluttered environment. One is that the radar range measurement of target is typically ambiguous, and in the absence of measurement noise, multiple possible target ranges project onto the same range measurements. Direct using of the range measurement for target state estimation may result in significantly poor estimate accuracy. The other challenges is that the target detection probability is usually imperfect and even low when the target is occluded, besides, lots of ground clutter measurements are received from the HPRF radar due to its look-down detection model.

Motivated by addressing the two main challenges described above, this thesis proposes a single-target multi-scan automatic target tracking algorithm in cluttered environment using the HPRF radar, i.e., the Gaussian mixture measurement (GMM)-integrated track splitting (ITS). In the GMM-ITS algorithm, numbers of unambiguous measurement components are created based on each ambiguous radar measurement due to the range ambiguity, the likelihood of each ambiguous radar measurement can be thus approximated by a mixture of set of weighted unambiguous measurement component likelihoods. In addition to addressing the range ambiguity problem, the GMM-ITS also recursively calculates the probability of target existence to provide a track quality measure in the cluttered environment for the false track discrimination (FTD), which is a track management procedure that recognizes and confirms true tracks, recognizes and terminates false tracks. Meanwhile, the existed multiple models (MM) algorithm for HPRF radar tracking provide a novel solution to resolve the range ambiguity problem by generating the pseudo-measurements with respect to each track model and obtain a convergent target state estimation, while the MM algorithm neglects the FTD for target tracking in clutter. Therefore, the MM is also enhanced in this thesis by integrating the random target existence event into the typical target state and recursively calculating the probability of target existence as an efficient track quality measure for the FTD. In the simulation study, the GMM-ITS algorithm shows improved FTD performance over the enhanced MM (EMM).

Furthermore, this thesis also extends the proposed GMM-ITS and EMM algorithms to a more practical multi-target tracking situation, and results in the GMM-joint ITS (GMM-JITS) and EMM-joint probabilistic data association (EMM-JPDA) algorithm, respectively. Both the GMM-JITS and EMM-JPDA are extended by employing the optimal Bayesian multi-target joint data association methodology to account for the situation when the targets move closely and the same measurements are selected by more than one track. The optimal Bayesian multi-target data association approach enumerates each feasible joint event which satisfies the constraint that each track is assigned to zero or one measurement and each measurement is allocated to zero or one track, and then evaluates their a posterior probabilities in order to calculate the a posterior data association probability and the probability of target existence for each track individually. Both algorithms are still capable of the FTD using the recursively calculated probability of target existence. The simulation study validates the efficiency of the extended algorithms, with the GMM-JITS shows superior performance in terms of the FTD and the track retention statistics.