The frequency of star centroid systematic error is analized, the common quality of star centroid systematic error distributions with different point spread functions is found. A simulation using Monte Carlo method is done according to the point spread function data of different field of views (FOV) from optical system. The result of simulation is in accordance with frequency analysis. The star centroid systematic error of the star object whose dispersion spot dimensions are 5 pixel × 5pixel is experimentally measured. The star centroid systematic error is compensated with error compensation equations, after compensation the accuracy of central FOV is increased by 66.56% , and the accuracy of fringe FOV is increased by 57.21% . So the effect of the compensation using the proposed method is better than the traditional compensation method using sine curve fitting with the accuracy only increased by 35.7%.