Aiming at the multi-target tracking with unknown clutter intensity, we proposed an Accelerated Expectation Maximization Probability Hypothesis Density (AEM-PHD) smoothing filter algorithm. Firstly, the model of clutter intensity was established, and the number of clutters was estimated according to clutter measurements. Then, the clutter density function was modeled by using Gaussian finite mixture model. AEM algorithm was proposed on the basis of EM algorithm, which was used for estimating the parameters of the Gaussian finite mixture model, and the clutter density function was obtained. Finally, the estimated clutter information was applied to multi-target tracking, and the target states were smoothed. Simulation results showed that, under clutter with unknown intensity, the proposed method can estimate clutter parameters accurately with high target tracking precision and accurate estimation of the target number.