A novel algorithm for defect extraction and classification of mobile phone screen based on machine vision

Abstract

Defect detection is a critical way for quality ensuring of mobile phone screens. In this paper, we propose a novel defect extraction and classification scheme for mobile phone screen based on machine vision. In order to improve the efficiency of the algorithm, a pre-examination algorithm and a coarse-precise defect extraction strategy are designed. Considering the problem that there are various types of mobile phone screen, a region of interest (ROI) acquisition algorithm is proposed to ensure the universality of the detection method. Besides, a clustering algorithm is proposed to avoid false detection or missed detection of cluster defects. Furthermore, the detection criteria are defined, and a classification algorithm combining multi-layer perceptron (MLP) and deep learning (DL) technologies is proposed. Experimental results demonstrate that satisfactory performance is achieved in detecting scratches, floaters, light stains and dark stains of the mobile phone screen with the proposed detection scheme.