随着现代航天技术的发展，航天器的应用效益的增值和在轨运营效率的提高就要求航天器要实现自主导航，而星敏感器是目前能够为航天器提供最高精度绝对姿态的测量器件。星图处理及质心定位精度的研究对星敏感器获得准确星点位置信息并进行姿态测量而言具有重要的意义。论文通过对星敏感器探测器上目标的真实分布情况进行分析，提出一种改进的形态学top-hat算法，能够有效地分割信号与背景噪声，提升图像质量。实验验证，该方法还适用于多尺度目标分类提取。分析星敏感器光学系统的点扩散函数（PSF）模型，借助傅里叶变换确定星敏感器入射光的截止频率，在满足奈奎斯特定理的条件下，利用高精度三维转台对星点进行亚像素级采样，还原星敏感器光学系统点扩散函数的高斯半径。从点扩散函数的对称性角度分析其对质心计算精度的影响，点扩散函数的对称性越高，则计算精度也越高。高斯半径大小决定质心计算窗口大小的选择，通常边缘噪声对质心计算精度影响较大，因此窗口不宜过大，但要尽可能囊括星点能量。提出变窗口质心定位方法，使质心定位精度相比于固定窗口提取方法提高30%。将改进的算法应用于带有云、太阳光、月亮等大背景杂光星图，根据星点与背景噪声的形貌区别，选择合适的结构元可有效提取出星点，满足星敏感器在大背景噪声下高精度姿态测量的需求。

With the development of modern aerospace technology, the value-added application of spacecraft and the improvement of on-orbit operation efficiency require spacecraft to achieve autonomous navigation. Star trackers are currently the measurement devices that can provide the highest precision absolute attitude for spacecraft. The study of star image processing and centroid positioning accuracy is of great significance for star sensors to obtain accurate star position information and perform attitude measurement.By analyzing the real distribution of the target on the star tracker detector, an improved morphological top-hat algorithm is proposed, which can effectively segment the signal and background noise and improve the image quality. Experimental verification shows that the method is also suitable for multi-scale targets extraction and classification.The point spread function (PSF) model of the star tracker optical system is analyzed. The cutoff frequency of the incident light of the star tracker is determined by means of Fourier transform. Under the condition of satisfying the Nyquist theorem, the star point is performed by high-precision three-dimensional turntable. Sub-pixel sampling, which restores the Gaussian radius of the point spread function of the star tracker optical system.The influence of the point spread function on the accuracy of centroid calculation is analyzed. The higher the symmetry of the point spread function is, the higher the calculation accuracy will be. The Gaussian radius determines the choice of the centroid calculation window size. Usually the edge noise has a great influence on the centroid calculation accuracy, so the window should not be too large, but the star point energy should be included as much as possible. A variable window centroid localization method is proposed to improve the centroid positioning accuracy by 30% compared to the fixed window extraction method.The improved algorithm is applied to images with large background stray light such as clouds, sunlight, moon, etc. According to the difference between the star point and the background noise, selecting the appropriate structural elements can be used to effectively extract the star points, satisfying the need for high-precision attitude measurement under large background noise.