In order to improve the efficiency and accuracy of on-orbit focusing of earth observation satellites, this paper proposes a method to quantitatively evaluate the defocusing state of the camera based on multiple star images and calculate the optimal detector position. First, the workflow of the auto-focusing algorithm is introduced. The calculation methods of star image preprocessing, image centroid and spatial phase difference (SPD) are described in detail. On the basis of SPD, the point spread function (PSF) is reconstructed and the standard deviation is calculated according to the discrete PSF fitting Gaussian curve. The standard deviation and the detector position are fitted as the focusing curve, and the lowest point of the curve is the optimal detector position calculated by the auto-focusing algorithm. Then, in order to illustrate the effectiveness of the algorithm proposed in this paper, the spatial domain, frequency domain, statistics and feature-based image sharpness evaluation function are briefly introduced as a comparison algorithm. Finally, a test platform is built to test the focusing accuracy and noise robustness of the proposed algorithm, and compared with other clarity evaluation functions. Experimental results indicate that the proposed algorithm achieves with a minimum error of -0.001 3 mm for the optimal detector position calculation, less than 1/6 semi-focus depth, better than the comparison algorithm ; when the image signal-to-noise ratio is not less than 10 dB, the absolute value of the focusing error is less than 0.002 8 mm, and the focusing curve fitting goodness is greater than 0.85 within the range of 1/6 semi-focus depth, and the noise robustness is better than the comparison algorithm. The proposed algorithm acquires multiple star images at different detector positions, quantitatively evaluates the camera′s defocus state through the star images, and the fitted focusing curve can correctly reflect the relationship between the camera′s defocus state and the amount of defocus. The calculation accuracy at the optimal detector position can meet the focusing accuracy requirements of space cameras and has strong noise robustness.