In dynamic background, in order to reduce the influence of background motion on target detection, the paper proposed a target detection algorithm based on motion vector compensation for video sequence images. In this paper, background motion compensation method is used to complete target detection. Firstly, the accurate matching points are obtained by using SURF feature points, and the accurate motion vectors are obtained; Then, the foreground and the background areas are distinguished according to the displacement of the moving target and the background motion vector, the background motion vector are substituted into the model to estimate the parameters, which can improve the accuracy of the background compensation; Finally, the moving target is detected by frame difference method. The experimental results demonstrate that the algorithm is robust and can detect moving objects accurately in the dynamic background.