<p>To address the issue of reduced tracking accuracy in close-proximity target scenarios caused by measurement-target association ambiguity in traditional extended target tracking (ETT) algorithms, this paper proposes a close-proximity extended target tracking algorithm that integrates morphological matching-based measurement set partitioning with a generalized labeled multi-Bernoulli sequential Monte Carlo (GLMB-SMC) filter. First, the target motion state predicted by the GLMB-SMC filter and the shape prior information generated by Gaussian Surface Fitting (GSF) are dynamically fed back to the measurement partitioning module. Next, the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm is used to perform an initial partitioning of the measurement set. Based on this, the prediction components with higher weights are used to judge the mixed measurement clusters of multiple neighboring targets. Combining the predicted target state and shape information, the fuzzy C-means (FCM) algorithm with the radial management strategy (RMS) is used to re-partition the mixed clusters to obtain more accurate results. Finally, the partitioning results are sent to the GLMB-SMC filter for updating. Simulation results demonstrate that the proposed measurement set partitioning method effectively utilizes the target position and shape information from the filter’s prediction step output to achieve precise partitioning of mixed measurement clusters, thereby enabling high-precision estimation of the kinematic state and extended shape of nearby extended targets.</p>