Rapid in-process 3D measurement of physically large components with complex shapes presents a challenging problem in modern industrial manufacturing. Although various robotic vision systems have been explored to provide the solution required for flexible and reconfigurable in-process 3D measurement, there is a key issue in its adoption due to the level of measurement accuracy that is achievable. Since this issue is directly linked to system calibration, it requires a rigorous error sensitivity analysis to ensure reliable and robust control of manufacturing processes. This forms the basis for the investigation reported in this paper. In particular, the paper focuses on practical and inherent measurement uncertainties in calibration of a robotic vision system. Starting from a brief review of popular approaches which have been proposed for hand-eye calibration between the robot and the camera, possible sources of uncertainties are identified and discussed. Based on a series of computer simulations carried out, the paper shows the impact of uncertainties on the results of hand-eye calibration for a robotic vision system, and provides the testing methodology as well as the statistical information required for selection of an appropriate calibration approach to achieve in-process 3D measurement with a higher confidence level.