Development of measurement system via deep learning based inspection system

To, Rui Xiang (2021) Development of measurement system via deep learning based inspection system. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

	Text (24 Pages) Development of measurement system via deep learning based inspection system.pdf - Submitted Version Download (545kB)
	Text (Full text) Development of measurement system via deep learning based inspection system.pdf - Submitted Version Restricted to Repository staff only Download (2MB)

Abstract

The measurements system is very vital part to the manufacturing industry. The dimension of electronic device is important and should be consistent. Automatic measuring instruments can be used to measure the size of electronic device such as IC chip, resistor and other. Compared to traditional method, automatic measurement can measure the size of an object fast and accuracy. In this project, the focus is on the development of measurement system via deep learning based inspection system. The major problem to development a deep learning model to detect the electronic item is hard, it requires a larger data set of images to achieve faster and high accuracy. The five electronic item that choose to detect are resistor, IC, button, led and PCB. The first objective of this project is to investigate image acquisition, image processing and illuminating system for measure the dimension of electronic item while the second objective is to development a measurement system via deep learning based inspection system by using python. Measurement system via deep learning based inspection system proposed in this project will be using laptop connected with smartphone camera. The camera will be fixed at a position so that the dimension of the electronic components can measure accurately. The program will be write using python language. The YOLO models that used to develop the system is YOLOv3, YOLOv4, tiny-YOLOv3 and tiny-YOLOv4. The YOLO model is train using the online GPU via Google Colaboratory. To test the functionality of measurement system, there two experiments are conducted. In the first experiment, the performances of YOLO models are analyzed. The accuracy of the measurement system via deep learning based inspection system in real-time is the second experiment. The measurement system via deep learning based inspection system is successfully developed.

Actions (login required)

View Item