The 3D laser scanning system mounted on an industrial robot is capable of detecting the surface information of parts quickly, and can enhance the system capability by improving the accuracy of measurement. In this paper, the measurement error of scanning system is analyzed and evaluated. The scanning system used in this experiment consists of an industrial robot and a commercial 3D laser scanner T-Scan. The basic principle of this commercial 3D laser scanner is laser triangulation, as a result, its measuring error is affected by scanning posture. In this experiment scanning posture is decomposed into scan depth, pitching angle and yawing angle. And the influence of scanning posture on random error and systematic error is studied by a series of experiments which control the variation of three parameters in sequence. Experiment result shows that the random error of the measurement is small compared with systematic error. And the systematic error shows a bilinear relationship with the scan depth and the pitching angle. A prediction model is developed and the maximum deviation between the model and the experimental results is only 26mm. The prediction model is a precondition to optimize the scanning trajectory and improve the measurement accuracy.