Rotational vision system (RVS) is a common type of active vision with only rotational freedom. Typically, the rotational freedom is provided by turntable and pan-tilt-zoom (PTZ). Or eye in hand (EIH) structure in an articulated arm robot. The ideal assumption that rotation axes are perfectly aligned with the coordinate axes of the local camera is mostly violated due to assembling deviations and limitations of manufacturing accuracy. To solve this problem, we propose a generalized deviation model for a specified rotation axis that relates the rotation motion of the platform to the exterior orientation (EO) of the camera. Based on it we put heuristic estimation algorithms through minimizing global reprojection error and fitting a circle in space respectively for rotating platform with or without accurate angle measurements with constrained global optimization. Implemented experiments on a servo pan-tilt turntable validate the accuracy and efficiency of the above models and calibration technique.