One method combining edge Gaussian mixture model and frame difference is proposed to realize the moving object detection under infrared light mutation, which uses the statistical threshold and image variance change to distinguish the occurrence of light mutation, and introduces a retard factor to upgrade the model. Experimental results show that, compared with the traditional detection method, the improved method here shows lower rate of false detection, more accurate detection, stronger robustness and stability, and it can effectively solve the false reporting problems induced by light mutation and infrared mode transition in the intelligent security monitoring system.