Computer Vision and Image Segmentation Implemented on GPU Using Compute Unified Device Architecture as Applied on Quality Inspection of Pre-etched Printed Circuit Board
Department of Computer Engineering of the University of San Carlos
World Congress on Engineering and Computer Science (WCECS), 2012
@inproceedings{astillo2012computer,
title={Computer Vision and Image Segmentation Implemented on GPU Using Compute Unified Device Architecture as Applied on Quality Inspection of Pre-etched Printed Circuit Board},
author={Astillo, P.V. and Patiluna, V.},
booktitle={Proceedings of the World Congress on Engineering and Computer Science},
volume={1},
year={2012}
}
Computer vision and image processing continue to expand its area of application. Traditionally, this technology was hosted by a sequential processing paradigm of a Central Processing Unit (CPU). With this implementation in mind limits the usefulness of a device that is capable of parallel processing for several years. At the same time, it has been observed that common problem encountered when image processing routines are rendered on CPU is a slow processing rate. This study presents the application of computer vision and image processing segmentation rendered on a Graphics Processing Unit (GPU), a parallel processing capable device, using CUDA developed by NVIDIA. It results to an impressive speed-up compared to the CPU. The study implements computer vision based quality inspection on pre-etched printed board fabricated by the Printed Circuit Board Prototyping Laboratory (PCBLab) of the University of San Carlos. The developed system can successfully detect defects such as open tracks, shorted tracks, neck form tracks, nick or mouse-bite form tracks, hole misalignment, and unwanted routes.
November 16, 2012 by hgpu
