For the infrared early warning system, the multi-target tracking algorithm usually has long starting time and poor real-time performance due to long measurement cycle and large quantity of targets. In this paper, a new algorithm based on random finite set is proposed. Considering that for the standard multi-target tracking algorithm based on the standard random set, it is necessary to give the prior birth probability density of the new target and the track can't be labeled, we make improvement to the algorithm. Sector processing method, uniform new target density model, and track batch management are used. Simulation results show that the improved algorithm can meet the demand of infrared warning system.