Abstract: This paper proposes a three-frame difference method based on adaptive learning rate Gaussian mixture modeling. Through the Gaussian mixture background modeling based on adaptive learning rate, the adaptive correction of the background model is realized, and the algorithm can completely extract the internal information of the target in the dynamic environment. Then, the three-frame difference improvement algorithm based on edge extraction is used to extract the target contour rapidly and use it as the edge complement. The experimental results show that the algorithm can completely extract moving targets and ensure the continuity and smoothness of the target edges. At the same time, the speed of target detection is increased, and it can be widely used in intelligent video surveillance, medical treatment and other fields.