Machine learning approach for automated optical inspection of electronic components

• Main Author: Lim, Siew Kee
• Format: Final Year Project / Dissertation / Thesis
• Published: 2019
• Subjects: T Technology (General); TK Electrical engineering. Electronics Nuclear engineering
• Online Access: http://eprints.utar.edu.my/3924/1/fyp_EE_2019_LSK.pdf; http://eprints.utar.edu.my/3924/

Rapid development in electronic industry causing a high volume of electronic components being manufactured on each day. Human visual inspection system has been traditionally used in electronic industry for quality control. However, human visual inspection system is affected by inconsistent between different inspectors, limitation of human ability, and is time consuming. (Smith and Adendorff, 1991). Human visual inspection system is insufficient for conducting quality control under such increased capacity. Thus, automated optical inspection system has been invented to replace human visual inspection system. Automated optical inspection system is a system that inspect the surface of the device under testing with the aid of machine vision under sufficient amount of light source. Automated optical inspection system become more crucial in the quality control area due to its inspection speed, inspection consistency and inspection accuracy. However, the technology used by automated optical inspection system keep on improving. This project proposed an automated optical inspection system implementing supervised machine learning algorithm and local features to inspect and classify defectiveness of surface mounted device light emitting diode into two classes: PASS and FAIL. PASS indicates that the product is a non-defective product while FAIL indicates that the product is a defective product. Supervised machine learning algorithm that work the best with the selected features for the inspection of surface mounted device light emitting diode is studied. The performance of the supervised machine learning is determined by the prediction accuracy. The factor that affecting the confidence level of the supervised machine learning algorithm is discussed.