A random matrix based method for tracking multiple extended targets

The paper presented an extension of random matrix-based approach [1] to track multiple extended targets in the cluttered environment, which can provide the estimates for target kinematics and target extension, as well as the corresponding identity information. To do this, the DBSCAN clustering (Density Based Spatial Clustering of Applications with Noise) method is employed to divide the whole measurement set into several distinct clusters, with each cluster representing the measurement subset from the same target or false alarm. Moreover, when multiple extended targets are in close proximity, it is hard to partition the measurement set accurately. In this case, the FCM (Fuzzy C means) method is incorporated to produce a soft clustering result. After that, inspired by the idea of Joint Probabilistic Data Association (JPDA), a series of operations are implemented to produce state estimates for multiple extended targets, including generating the feasible events, evaluating the association probabilities and integrating the estimates from multiple association hypotheses, etc. Simulation results demonstrated the effectiveness of the proposed approach.