The typical applications of adaptive background prediction technology in weak and small infrared target detection are summarized in this paper. The advantages and disadvangtages of linear and nonlinear adaptive algorithms are analyzed. The traditional LMS algorithm is improved. The improved algorithm processes the consecutive frames according to the motion characteristics of targets. The experimental results have shown that this algorithm can effectively suppress the false alarm caused by the edge of cloud and its capability to process unstable and nonlinear clutter background and low SNR images is improved.