Development of Adaptive and Intelligent Software Components for SMT Automatic Optical Inspection Systems
碩士 === 長庚大學 === 機械工程研究所 === 92 === This thesis aims to use the image processing and artificial neural network (ANN) to develop the automatic optical inspection (AOI) system to classify and recognize the 2D solder paste defects and to reconstruct 3D virtual laser solder paste surfaces. Mea...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | en_US |
Published: |
2004
|
Online Access: | http://ndltd.ncl.edu.tw/handle/22871945602950783975 |
id |
ndltd-TW-092CGU00489003 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-TW-092CGU004890032016-01-04T04:08:37Z http://ndltd.ncl.edu.tw/handle/22871945602950783975 Development of Adaptive and Intelligent Software Components for SMT Automatic Optical Inspection Systems SMT機台自動光學檢測系統之適應式智慧型軟體元件開發 Fang-Chung Yang 楊方中 碩士 長庚大學 機械工程研究所 92 This thesis aims to use the image processing and artificial neural network (ANN) to develop the automatic optical inspection (AOI) system to classify and recognize the 2D solder paste defects and to reconstruct 3D virtual laser solder paste surfaces. Meanwhile, we propose the pad-based and sub-area based methodology cooperating with the in-lab design specified light projection source simulation system and commercial coaxial light source to develop the AOI system. The back-propagation ANN module is also developed for training and recall inspection patterns. With the in-lab design structure light source, several AOI images are acquired to reconstruct the 3D solder paste surface model, however, it consumes a lot of time for images acquisition and computing. On the other hand, the commercial coaxial light source system is adopted to capture images for reconstructing the virtual laser 3D surface model. With the experimental results, the in-lab design structure light source for 3D surface reconstruction achieves 95% accuracy rate; In the same time, the commercial coaxial light source for 3D reconstruction just achieves about 80% recognition accuracy, nevertheless, it already meets the requirements of the industry applications. In summary, the 2D solder paste defect inspection achieves 85% average recognition accuracy; and the 3D solder paste surface reconstruction achieves 90% volumetric accuracy when compared to the actual laser surface scanning. In additional, the proposed system had been implemented as a window-based application program using the Microsoft Visual C++. The modules include the image acquisition, image processing, back-propagation ANN training and recall functions, 2D solder paste defects identification and 3D solder paste surface reconstruction, and graphical display interface. They are justified to most the SMT AOI inspection requirements. Chung-Hsien Kuo 郭重顯 2004 學位論文 ; thesis 85 en_US |
collection |
NDLTD |
language |
en_US |
format |
Others
|
sources |
NDLTD |
description |
碩士 === 長庚大學 === 機械工程研究所 === 92 === This thesis aims to use the image processing and artificial neural network (ANN) to develop the automatic optical inspection (AOI) system to classify and recognize the 2D solder paste defects and to reconstruct 3D virtual laser solder paste surfaces. Meanwhile, we propose the pad-based and sub-area based methodology cooperating with the in-lab design specified light projection source simulation system and commercial coaxial light source to develop the AOI system. The back-propagation ANN module is also developed for training and recall inspection patterns. With the in-lab design structure light source, several AOI images are acquired to reconstruct the 3D solder paste surface model, however, it consumes a lot of time for images acquisition and computing. On the other hand, the commercial coaxial light source system is adopted to capture images for reconstructing the virtual laser 3D surface model. With the experimental results, the in-lab design structure light source for 3D surface reconstruction achieves 95% accuracy rate; In the same time, the commercial coaxial light source for 3D reconstruction just achieves about 80% recognition accuracy, nevertheless, it already meets the requirements of the industry applications. In summary, the 2D solder paste defect inspection achieves 85% average recognition accuracy; and the 3D solder paste surface reconstruction achieves 90% volumetric accuracy when compared to the actual laser surface scanning. In additional, the proposed system had been implemented as a window-based application program using the Microsoft Visual C++. The modules include the image acquisition, image processing, back-propagation ANN training and recall functions, 2D solder paste defects identification and 3D solder paste surface reconstruction, and graphical display interface. They are justified to most the SMT AOI inspection requirements.
|
author2 |
Chung-Hsien Kuo |
author_facet |
Chung-Hsien Kuo Fang-Chung Yang 楊方中 |
author |
Fang-Chung Yang 楊方中 |
spellingShingle |
Fang-Chung Yang 楊方中 Development of Adaptive and Intelligent Software Components for SMT Automatic Optical Inspection Systems |
author_sort |
Fang-Chung Yang |
title |
Development of Adaptive and Intelligent Software Components for SMT Automatic Optical Inspection Systems |
title_short |
Development of Adaptive and Intelligent Software Components for SMT Automatic Optical Inspection Systems |
title_full |
Development of Adaptive and Intelligent Software Components for SMT Automatic Optical Inspection Systems |
title_fullStr |
Development of Adaptive and Intelligent Software Components for SMT Automatic Optical Inspection Systems |
title_full_unstemmed |
Development of Adaptive and Intelligent Software Components for SMT Automatic Optical Inspection Systems |
title_sort |
development of adaptive and intelligent software components for smt automatic optical inspection systems |
publishDate |
2004 |
url |
http://ndltd.ncl.edu.tw/handle/22871945602950783975 |
work_keys_str_mv |
AT fangchungyang developmentofadaptiveandintelligentsoftwarecomponentsforsmtautomaticopticalinspectionsystems AT yángfāngzhōng developmentofadaptiveandintelligentsoftwarecomponentsforsmtautomaticopticalinspectionsystems AT fangchungyang smtjītáizìdòngguāngxuéjiǎncèxìtǒngzhīshìyīngshìzhìhuìxíngruǎntǐyuánjiànkāifā AT yángfāngzhōng smtjītáizìdòngguāngxuéjiǎncèxìtǒngzhīshìyīngshìzhìhuìxíngruǎntǐyuánjiànkāifā |
_version_ |
1718158496091865088 |