Non-rigid Extended Target Tracking Based on STGP-ETCBMeMBer Filter

‍ ‍Shape estimation is one of the difficult aspects of extended target tracking， and inaccurate modeling methods can lead to poor estimation results. For the tracking problem of irregular multi-extended targets， a Cardinality Balanced Multi-Extended Target Multi-Bernoulli （STGP-ETCBMeMBer） filtering algorithm is proposed based on the Spatio-Temporal Gaussian Process （STGP） model. First， the augmented multi-extended target state set and the measurement set are modeled as multi-Bernoulli random finite sets and Poisson random finite sets， respectively， using the random finite set method， and on this basis， the STGP method is used to model the measurement sources of the star-convex extended target to improve the algorithm’s shape estimation accuracy of the extended target. After that， in the algorithm update stage， multiple likelihood functions corresponding to the same target measurement subset are assumed to obey Gaussian distribution， and closed form solution realized by Gaussian mixture （GM） implementation is derived. Finally， the effectiveness of the proposed algorithm is verified by constructing simulation comparison experiments， and the simulation results show that the proposed algorithm has a more accurate effect on the shape estimation of extended targets.