Image processing and image analysis of defects in microlens array using smartphone camera and application

• Main Author: Teoh, Yuann Yun
• Format: Final Year Project / Dissertation / Thesis
• Published: 2022
• Subjects: T Technology (General); TA Engineering (General). Civil engineering (General); TK Electrical engineering. Electronics Nuclear engineering; TR Photography
• Online Access: http://eprints.utar.edu.my/4815/1/fyp_EE_TYY_2022.pdf; http://eprints.utar.edu.my/4815/

Microelectronics is crucial to the advancement of technology. However, industrial inspection of micro-devices is a very difficult and time-consuming operation, particularly when those devices are manufactured in large quantities utilising micro fabrication techniques. According to Marie Freebody, the contributing editor of Photonics Media, many optical surfaces are still examined with human visual inspection system which is insufficient for getting a more precise defects inspection when the objects are small in size. Thus, automated optics inspection (AOI) machine is invented to calibrate on optics surfaces and inspect for scratches. However, according to the quotation summarized by VCTA, a manufacturer in AOI, an AOI system is expensive with a cost of around 30,000RMB to 50,000RMB in China. In this project, a smartphone camera model that imitates a light field camera (LFC) equipped with a micro-lens array (MLA), a laser and a diffuser is developed. The phone camera is calibrated with camera calibration system. Then, the MLA is positioned in front of the laser, turning it into an array of tiny cameras that emits light field. The images produced are then processed on an application on a PC which implemented image processing method to classify the defectiveness of MLA with two classes: PASS or FAIL. Additionally, the images produced are also processed on Siamese Neural Network to find image similarity and classify the defectiveness with two classes: 0 or 1. With the proposed method, the intensity distributions, contour plot, 3D surface plot, training total loss on test MLA datasets are determined, discussed and studied.