A Maneuvering Non-ellipsoidal Extended Target Tracking Algorithm Based on PMBM Filter

‍ ‍The shapes of extended targets are mostly irregular and changes with different observation angles. In addition， there is problem of tracking maneuvering targets in complex scenarios of coexisting point and extended targets. To solve these problems， a maneuvering non-ellipsoidal extended target tracking algorithm based on Poisson multi-Bernoulli mixture filter is proposed. Firstly， the multiple sub-objects gamma Gaussian inverse Wishart （GGIW） represented by a sub-random matrix is used to replace the single GGIW density representing the extended target in the Poisson multi-Bernoulli mixture （PMBM） filter for improving the shape estimation accuracy. Secondly， when the target motion state changes， the shape information of the extended target usually changes as well， and consequently， the performance of multiple target tracking algorithms based on the constant number of sub-objects will decrease. Thus， DBSCAN algorithm based on density is employed for measurement clustering， achieving time-varying shape estimation of extended target by calculating the existence probability of sub-target components， thereby adaptively adjust the number of sub-targets. Finally， the algorithm achieves simultaneous tracking for coexisting multiple point and maneuvering extended targets by integrating interactive multiple model algorithm. Experimental results show that the proposed algorithm can effectively track maneuvering multiple extended targets with a good tracking performance for shape changes.