Abstract: Small target detection in complex background is a critical technology in the infrared warning system. Aiming at the problems that there is clutter interference in infrared image and the target-s signal-to-noise ratio is low, this paper presents a non-linear spatial filtering detection method. Based on traditional linear spatial filter algorithm, the algorithm calculates the gray value around the four quadrants of the potential target, and adjusts the dynamic threshold properly. The results show that when the background contains more complex factors, the non-linear spatial filtering method can effectively suppress the clutter to achieve the extraction of the weak target. Compared with the results of linear filtering algorithm, this algorithm decreases the number of false alarms by 3/4, and has easy engineering realization.