To solve the limitation in the gaze tracking system using active infrared light, a gaze tracking system under the condition of natural light allowing the human head moves freely is constructed. Binocular stereo vision method is employed to obtain the characteristic parameters of eyes, and then the sight line can be estimated. The head pose can be estimated by using headset calibration board and the error caused by head moving can be reduced. Then a calibration method of individual parameters in gaze tracking system is proposed. This method assumes the optical axis is coincided with the line of sight. Then the center and radius of eyeball, the angle between eye optical axis and visual axis can be calculated. The optimal parameters can be obtained by employing the nonlinear optimization algorithm to process the initial estimate values. Experiment results verify that the proposed method is almost unconstrained, and the estimation accuracy of sight line is high. So this method can satisfy the need of gaze tracking system in many cases.