In this paper

we propose a small target tracking method based on duo-temporal-scale signal analysis (DTS).The proposed tracking method exploits the natural difference between objects and noise in two phases:short-temporal-scale noise reduction phase (STS) and long-temporal-scale motion analysis phase (LTS)

which simulates the small target tracking process in human vision system.In STS

we propose a new space-temporal noise filter to remove random noise or blocks that are easy to be distinguished in appearance

and then we get the candidate targets with high probability to initialize moving targets' trajectories.In LTS

we propose a new trajectory smoothness constraint function based on different motion patterns between small targets and noise.Finally we get the small targets' trajectories by optimizing smoothness constraint function in long temporal scale using Viterbi algorithm.Experiments on different kinds of outdoor scenarios show the efficiency and robustness of the proposed method.