Aiming at the problems of the existing box particle probability hypothesis density (PHD) based group target tracking algorithm such as a heavy computational burden, poor stability of the extracting state with a large number of clusters and unavailability of group trajectories, a labeled box particle PHD group tracking algorithm is proposed. First, the measurements are preprocessed to eliminate the clutter measurements, so as to reduce the computational burden of the measurement updating step. Then, by adding labels to the box particles, different group targets are distinguished, and the trajectories of different group targets can be obtained. Finally, the states of group targets are extracted according to different labels and the impact of k-means clustering instability are effectively avoided. Simulation experiments illustrate the advantages of the proposed algorithm in terms of light computational burden, track maintenance of different groups under the environment of miss detection, and accurate extraction of the group target state in the case of a large number of clusters.