In the space-based space target monitoring system, stray light in the space environment causes very serious non-uniform distribution characteristics in the background of the star map, which greatly affects the accuracy in extracting star. Therefore, it is necessary to preprocess star images under stray light background. Traditional preprocessing methods have difficulty in dealing with complex stray light backgrounds, leading to poor star extraction performance. To address the issue, this paper introduces a multi-scale grayscale morphology method for star extraction under stray light background. First, median filtering is performed on the image to effectively suppress high-frequency noise and impulse noise and smooth the non-uniform background. Then, the image is further processed using dual-structure multi-scale grayscale morphological transformation to effectively suppress background noise and enhance the contrast of the target area. Then, adaptive threshold processing is performed to accurately separate star and background areas under different brightness conditions. Finally, the stars are extracted through connected domain analysis. The results of the semi-physical simulation experiment prove that the peak signal-to-noise ratio PSNR is 39 dB, the background suppression factor BSF is 5.729e-4, and the average running time of the method is 0.126 s. This method can greatly suppress background noise and greatly improve the accuracy of star extraction, enabling a higher processing speed.