| Summary: | 碩士 === 國立臺灣科技大學 === 自動化及控制研究所 === 102 === This paper developed an optical inspection system of CMOS image sensor (CIS) micro multi-layer non-spherical lens module. However, the CIS lens module structure was multi-layer. When light passed through the non-spherical lens module, it will cause a halo because of reflection and refraction. It led defect is difficult to be found. When acquiring the image, the image resolution and size of defects needed to be considered. It caused the magnification lens’s depth of field does not cover the height of the inspected object. The mentioned two cases would be difficult to perform a follow-up defect detection so this paper developed solutions these problem, and developed a CIS micro multi-layer non-spherical lens module inspection system.
The CIS micro multi-layer non-spherical lens module inspection system consisted three main components: (1) Image preprocessing, (2) Defect inspection procedure, and (3) Defect recognition. The first component mainly used Hough transform to detect circle and to segment region of interest (ROI), and then amended illumination unevenness by using single-scale Retinex. The second component used the Kuwahara filter, control limit method, region growing, etc to segment the defect. In order to solve the problem of lens’s depth field not covering the inspected object’s height, the multi-image was acquired. This paper proposed a way to avoid major defect that influenced different layers image by judging defect unabridged rate and defect sharp rate, and calculated the defect feature. Finally, this study used a support vector machine to recognize different kinds of defect. The experimental result showed the accuracy of defect recognition is 97.97 %. This paper proposed an optical inspection system of CMOS image sensor micro multi-layer non-spherical lens module that effectively detected and recognized the defects.
|
|---|
