A temporal/spatial fusion filtering algorithm was proposed for detecting small infrared target. The algorithm suppresses most of background and enhances targets by Top-hat filter in spatial, and suppresses high background profile and most of random noise by three frames difference filter (TFDF) in temporal. There are only target and little noise in the image after temporal/spatial filtering. Then an or-operation was proposed to accumulate energy of sequence image. Target track can be detected after or operation. Morphological closing operation was applied to joint ruptured target track. At last, the adaptive threshold was adopted for detecting target track. The experimental results show that low SNR target with low speed and high speed can be detected in complex background.