Calibration is the estimation of the internal parameters of star sensor,which can decrease the system error and improve the pointing precision. Based on the different calibration method between coefficients to be determined and internal parameters,the calibration models are established. The different approaches of ground calibration and data processing are summarized according to different calibration models. The on orbit calibration method of focal length and optical center offset is analyzed,and the on orbit star images are processed with extended Kalman filter. The simulation results indicate that the new method can converge more quickly and eliminate the random centroiding measurement errors more efficiently than the nonlinear least squares estimation.