Aiming at the difficulty of infrared target detection of “low and slow small” unmanned aerial vehicles (UAV) in complex low-altitude background, this paper proposes a new target detection algorithm based on multiscale fusion filtering. Combined with spatial multiscale decomposition filtering and temporal multiscale difference processing, the algorithm can effectively overcome many difficulties such as complex Iow-altitude background interference, unknown target scale, unknown angular velocity and low target signal-to-noise ratio (SNR). The test result shows that the algorithm can effectively detect the UAV targets with different distances in complex Iow-altitude background, and the false alarm rate is low. The algorithm is realized in TI 6657 DSP and realizes 100Hz real-time processing of mid-wave infrared images with 640*512 resolution, which has been effectively applied to the large-field circumferential scanning infrared search system developed by ATR Lab.