Defect detection on air bearing surface with luminance intensity invariance

One important part of a read/write head in hard disk drive is Air Bearing Surface (ABS), therefore its verifying is necessary. Main problem of verifying ABS with machine vision in manufacturing process is variance of luminance intensity that affects image acquisition. This research proposes method for reduce variance of luminance intensity by Texture unit, then extracting the features of ABS with Principle Component Analysis, and selected threshold for clustering defected and non-defected ABS by mean of threshold set which has minimum summation of two types error as false rejection and false acceptance from training data. The experimental results show that without Texture unit, defected and non-defected ABS cannot be clustered. Thus proposed method is better used because it can be reduce variance of luminance intensity.