A refined track initiation algorithm is proposed for the objects in dense formation. The measurement clustering algorithm, which needs significant amount of computation, is not needed in the method. First, all possible associate point sequences are built up by extending the track initiation conditions. Then, clustering is made to all the sequences by comparing the different sequences to see if they have the common measurements. At last, optimized distribution is taken to the initial relative sequences belong to the same cluster, and the optimum scheme is selected to each object in the formation for track initiation. Simulation results prove that the method has fine performance on formation track initiation, low computational burden and high accuracy rate on track initiation.