An extended target tracking algorithm based on trajectory feedback constrained spatial information

Accurately estimating the state of extended targets is a major challenge because measurement numbers and distributions change significantly, especially when tracking targets are close or overlapping. To solve the track fragmentation problem caused by spatial measurement ambiguity, a novel Trajectory Situation Feedback-based Gaussian Mixture Model Expectation Maximization (TSF-GMM-EM) method is proposed within the framework of the Gaussian Mixture Probability Hypothesis Density (GM-PHD) filter. The proposed method feeds historical trajectory information back to guide current measurement partitioning and motion estimation while providing prior-guided initialization for the EM process, thereby improving partitioning reliability during target intersections and accelerating clustering convergence. Experimental results show that the TSF-GMM-EM method limits the peak OSPA error to approximately 12 m during target intersections. It also maintains a measurement partitioning accuracy above 0.946 in overlapping target scenarios.