This study is related to automation of polishing process by using an industrial robot. In polishing process, full automation of polishing operation has not been achieved due to the difficulty in visual inspection. Judgment of polishing sufficiency has been performed by skilled workers. In this study, the surface after milling process is the target for polishing. In the previous report, the generation of robot program for polishing operation based on CAD data was already proposed. Focusing on the periodicity of cutter mark patterns on milling surface, we proposed an automatic judgment of polished surface quality. This method is based on the image processing which uses Discrete Fourier Transformation. In order to capture images of the polished surface, a CCD camera was mounted on a 6-axis controlled robot hand. However, the image processing method proposed in previous report could not judge the curved cutter mark patterns because the periodicity of those patterns is not uniform. The objective of this report is the development of automatic judgment method for non-periodic cutter mark patterns. We developed a judgment method which uses statistical pattern recognition as a means to judge the existence of cutter marks. As a result, a system that allows the automatic polishing that includes the surface quality judgment is developed.