Automated defect inspection system for CMOS image sensor with micro multi-layer non-spherical lens module. (October 2017)

• Record Type: Journal Article
• Authors: Jeffrey Kuo, Chung-Feng; Lo, Wen-Chi; Huang, Yan-Ru; Tsai, Hsin-Yang; Lee, Chi-Lung; Wu, Han-Cheng
• Is Part Of: Journal of manufacturing systems. Volume 45(2017)
• Page Start: 248
• Page End: 259
• Publication Date: 2017-10
• Subjects: Complementary metal oxide semiconductor -- Automatic inspection -- Image processing algorithm -- Micro multi-layer aspherical lens module; Manufacturing processes -- Periodicals; Production engineering -- Data processing -- Periodicals; Robots, Industrial -- Periodicals; Production, Technique de la -- Informatique -- Périodiques; Robots industriels -- Périodiques; Electronic journals; 670.42
• DOI: 10.1016/j.jmsy.2017.10.004
• Languages: English
• ISSNs: 0278-6125

Abstract

Highlights: Used the Retinex algorithm to correct the brightness of the multilayer lens module. Generated a state-of-the-art experimental machine to solve the depth of field. Utilized the defect integrity and defect sharpness to avoid extensive defects. Compared single-scale Retinex with multiple-scale Retinex for brightness correction. Abstract: This study presented an optical inspection system for micro multi-layer non-spherical lens module of complementary metal-oxide-semiconductor (CMOS) image sensor. The lens module structure was multi-layer. When light passes through the non-spherical lens module, it causes a halo due to reflection and refraction, and leads to defects that are difficult to detect. When extracting the image, the image resolution and size of defects need to be considered. The depth of field of the magnification lens does not cover the height of the inspected object. These two problems make follow-up defect detection difficult. This study provided an innovative solution to deal with these issues, and developed a CMOS image sensor (CIS) with non-spherical lens module inspection system. The proposed design consisted of three stages: 1) image registration and brightness correction: applying Retinex algorithm for brightness correction, in order to avoid difficulties in subsequent defect detection resulted from the light source into multilayer lens module; 2) defect detection: using image processing to segment the defects in the analyte, implementing an experimental machine to solve the depth of field that could not cover the analyte height, and performing the Z-axis mobile platform to capture multilayer image; to avoid extensive defects influencing the imaging quality of nearby layers, using the defect integrity and defect sharpness for identification, thus eliminating the effect of extensive defects, segmenting the defects accurately, and calculating the feature values; 3) defect identification: applying the support vector machine (SVM) for classification. The experimental result showed that the accuracy of defect recognition was more than 94.7%. The presented hardware architecture and on-line software processing procedure can be applied in the real inspection system.