In order to detect small moving targets effectively from the low SNR infrared image sequence under the interference of cloud background and noise,morphology Tophat filter was used to enhance the target by filtering out the low frequency background.Then,the Gaussian mixture background model was used for background estimation and adaptive background updating,to separate the foreground part from the background.Experimental results showed that:The Tophat filtering used together with Gaussian mixture background model can detect infrared targets accurately from the dynamic cloud background.