Surface defect inspection and classification for glass screen of mobile phone

Complex manufacturing makes the screens one or more defects simultaneously. Effective detection of defects is crucial in the manufacturing processes of mobile phone glass screens. In this paper, we concentrate on extracting and classifying some typical defects, such as scratch, bruise, pit and blister. Firstly, we use morphological filter to smooth background noise, and an improved gamma grayscale transformation is proposed to enhance the contrast. Then, a double threshold segmentation algorithm, based on area threshold and gray threshold, is presented to extract defects from background. Finally, according to different optimal segmentation feature values of different defects, a binary tree classifier is constructed to classify defects. The experiment results show that the proposed method can extract and recognize typical defects precisely.