A novel calibration method for multiple FOV star sensors based on three axis turntable was proposed. The method mainly took advantages of the turntable′s three rotational freedom degrees to calibrate FOVs of arbitrary directions without reinstalling the sensor. Modeling for laboratory calibration was achieved through optimizing and trimming of the observation model, the structure model and the external parameter model. The observation model parameters of each FOV and the structure model parameters among distinct FOVs were solved by the Levenberg-Marquardt nonlinear least square algorithm. Without the need of outfield star observation in the calibration process of structure model parameters, huge amount of data sampling work load was saved, hence, the estimation error caused by the atmosphere refraction and disturbance phenomena was avoided. The validity of the method was demonstrated by the simulation of a triple FOV digital star sensor and the real experiment of a dual FOV star sensor. Compared to the conventional method that utilizes outfield star observation data, the average angle distance error within single FOV reduces by 20.32%, and the average angle distance error between FOVs reduces by 59.34%.