Abstract: Satellite video stabilization is the prerequisite and basis for realizing high-precision satellite video applications.Due to the insufficient accuracy of satellite pointing attitude and the insufficient stability of platform attitude，image alignment technology is usually implemented to achieve the video gaze effect; however，when observing the sea targets，there is no control point calibration and feature point-based alignment cannot be carried out between frames，the staring video camera has the problem of repeated jumping of the targets on the image plane during observation.In this paper，a global foreground video stabilization （GFVS）method is proposed based on multi-target ship detection at sea.A Gaussian error model is constructed，and the image plane is corrected with the deviation between the optimized position and the original position，and finally the stabilized video is synthesized.The experimental results show that the method in this paper can effectively solve the problem of difficult alignment of dithered images when the control points at sea are insufficient，and get more stable video effects，the error of the stabilization can be controlled within 0.9 pixels by applying two datasets collected by the Jilin-1 satellite constellation.

Key words: maritime gaze observation, global foreground video stabilization, multi-objective optimization methods, stabilize images without feature points, Gaussian error model