Interacting multiple model probabilistic data association filter using random matrices for extended target tracking

Özpak, Ezgi
In this thesis, an Interacting Multiple Model – Probabilistic Data Association (IMM-PDA) filter for tracking extended targets using random matrices is proposed. Unlike the extended target trackers in the literature which use multiple alternative partitionings/clusterings of the set of measurements, the algorithm proposed here considers a single partitioning/clustering of the measurement data which makes it suitable for applications with low computational resources. When the IMM-PDA filter uses clustered measurements, a predictive likelihood function for the extent measurements is necessary for hypothesis probability calculation. Alternative predictive likelihood functions proposed in the literature for this purpose are surveyed and their shortcomings are identified in the thesis. Then an alternative predictive likelihood function is proposed and its advantage is illustrated on simulations running IMM-PDA filters with different predictive likelihood functions on a scenario involving a fighter aircraft launching a missile. When a single partitioning/clustering is used before the tracking operation as is the case for the tracker proposed in the thesis, the clusters corresponding to close targets might be merged by the pre-clustering step, which might lead to track loss in the tracker. For overcoming this problem, a specific algorithm is proposed for handling close targets and its performance is illustrated on a simple scenario.