An Estimation of the Quality of Hairline Finished Surfaces by Applying Textural Analysis

Hairline finishing, which leaves a detailed continuous line pattern on the surface of a metal, is being used widely in industrial products for its ornamental and functional effects. However, the evaluation of their quality is done, in many cases, sensuously and subjectively though visual inspection by skilled workers because evaluation is based on geometrical continuity and regularity of the line pattern and other kinds of information. This research aims at developing a way to evaluate hairline finished surfaces objectively and quantitatively by the image processing. First, the pictures of hairline finished surfaces were taken using a CCD camera, and the textural analysis was carried out by estimating the gray level co-occurrence matrix and the gray level run length matrix. This textural analysis is effective because there was a clear difference in textural features between the pictures judged as good and the pictures judged as bad. In consideration of these results, the evaluation system using the neural network was developed.