Effective defect classification for flat display panel film images

Abstract

In this paper, we propose an effective defect classification system for FDP (flat display panel) film images which are acquired in real production lines. A film image is segmented into a binary image with two non-overlapping regions: defect and non-defect regions. From the defect regions, various features are extracted such as brightness distribution, linearity, and morphologic characteristics. The film defects are classified through the analysis of those features extracted. Empirical study shows our system classifies five types of film defects effectively.