Automated visual inspection of magnetic disk media

This paper presents a novel approach to automating the visual inspection of magnetic disks. In commercial production, quality control is currently achieved by means of functional tests. However, due to increasing storage densities these methods are becoming obsolete. Automated visual inspection of the disk surface is required to achieve improved reliability. In this paper the defect classification stage of the inspection system is presented. Suitable methods of imaging and image processing are presented. In particular a novel method of texture recognition, based on n-tuple pattern recognition, is presented as a computationally efficient method of defect classification. The performance of this novel method is first compared with existing texture algorithms using the Brodatz texture album before preliminary results are shown for some frequently occurring disk faults.