Development of product flaw detection algorithm based on deep learning

Tan, Yong Sing (2021) Development of product flaw detection algorithm based on deep learning. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

	Text (24 Pages) Development of product flaw detection algorithm based on deep learning.pdf - Submitted Version Download (589kB)
	Text (Full text) Development of product flaw detection algorithm based on deep learning.pdf - Submitted Version Restricted to Repository staff only Download (2MB)

Abstract

In recent years, technology is involving rapidly and PCBs are very crucial component in most electronic devices, thus, demands of PCBs are gradually increasing days by days. Therefore, PCBs inspection processes have to continually improve to meet the increasing demand of PCBs. Besides that, PCBs are now having more and more complex designs and tinier components, therefore, human inspection system is not suitable to meet the high accuracy and high-speed inspection. So, industries have attempted to achieve nearly 100% quality inspection system by using machine vision to ensure the quality of all PCBs with also the help of new era of inspection method which is deep learning algorithm. However, inspection method with deep learning algorithm has not been widely implemented yet. In this project, the focus is on the design and development of a product flaw detection using deep learning algorithm. This is used to detect components missing and distance detection on the PCBs. The hardware components are built by a Sony XZs smartphone camera, LED system and laptop with GPU processor. The software is using Python language, PyCharm IDE, OpenCV library, different types of YOLO deep learning algorithm. The YOLO deep learning object detection networks are trained using Google Colab GPU. Experiments have been conducted to compare the accuracy between all four type of YOLO algorithms with different lighting conditions.

Actions (login required)

View Item