In order to obtain optimization threshold for the centroid algorithm,several kinds of influencing factors,such as signal-to-noise ratio,discrete gathering of star point and so on,are studied.Based on theoretical computation,a function is established to depict the relationship between the luminosity of star spot and the threshold with optimization centroid accuracy.Then,by combining the function with the simulation results of discrete gathering of star spot,a method of obtaining the optimization threshold based on luminosity of star spot is proposed.Based on the relationship between optimization threshold and luminosity of the spot,a curve model is established,and the optimization threshold is educed by interpolation.The threshold estimation accuracy of the method is shown,and the centroid accuracy is also compared with those of various thresholding algorithms.The experimental results indicate that the luminosity estimation error of the optimization threshold is less than 3 under different SNR conditions of the camera system,and the centroid locating accuracy by proposed method has raised by 1/100～1/10 pixel level as compared with those of other methods,which can satisfy the requirement of dynamic threshold optimization of centroid algorithm.