A Novel Defect Detection Algorithm for Flexible Integrated Circuit Package Substrates

Abstract

Efficient fabrication and high reliability of flexible integrated circuit package substrates (FICS) urgently need effective defect detection methods. Based on the defect categories of FICS images, including yellow circuit backgrounds, film backgrounds with speckles, and black backgrounds, a novel rapid detection algorithm of impurity defects in FICS images is proposed. Three successive procedures are indispensable to realize the rapid detection with high accuracy, including image contrast enhancement based on image gray value analysis, obtaining the standard templates to weaken the texture interferences, and the probability calculation of the background feature of each pixel in the edge region between background and defect. Then, the accurate identification of film/black backgrounds and impurity defects in high-density FICS images is achieved. Furthermore, the comparisons of time–accuracy tradeoff obtained by using existing methods (Lk-means, Ostu, GMEM, SPCNN, MBGCT, FTC) and the proposed algorithm are carried out to experimentally verify the feasibility. The defect detection algorithm will facilitate a solution for the high similarity between the inherent texture features of FICS and external impurity defects at micrometer scale and substantially improve the detection accuracy and efficiency of high-density FICS, which will meet the real-time industrial requirement for the rapid fabrication of high-density FICS.