Automatic Surface Inspection Using 1-D Gabor Filters
碩士 === 元智大學 === 工業工程研究所 === 89 === In this study, we use machine vision to defect embedded in homogenously textured surfaces. In order to avoid noise interference in the spatial domain, we employ the Gabor transform method in the spatial-frequency domain to detect local defects. Tradition...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | zh-TW |
Published: |
2001
|
Online Access: | http://ndltd.ncl.edu.tw/handle/74743069775361524256 |
id |
ndltd-TW-089YZU00030035 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-TW-089YZU000300352015-10-13T12:14:43Z http://ndltd.ncl.edu.tw/handle/74743069775361524256 Automatic Surface Inspection Using 1-D Gabor Filters 一維賈柏轉換之表面瑕疵檢測 Chih-Ping Lin 林志賓 碩士 元智大學 工業工程研究所 89 In this study, we use machine vision to defect embedded in homogenously textured surfaces. In order to avoid noise interference in the spatial domain, we employ the Gabor transform method in the spatial-frequency domain to detect local defects. Traditional Gabor-based methods use 2-D Gabor filters for texture analysis. They are computationally intensive and affected by rotation. Given a problem with image size and filter size , the computational complexity of 2-D Gabor filters is . The proposed method in this study first 1-D ring-projection transformation to compress 2-D images to 1-D signals, and then employs 1-D gabor filters to detect defects. In this way, the computational complexity can be significantly reduced to , and the detection result is invariant to rotation. Both structural textures such as machined surfaces and textile fabrics and stochastic textures such as leather and castings in gray-level and color image are investigated. Experimental results have shown that the proposed method is effective and efficient for detecting local defects in textured surfaces. Du-Ming Tsai 蔡篤銘 2001 學位論文 ; thesis 130 zh-TW |
collection |
NDLTD |
language |
zh-TW |
format |
Others
|
sources |
NDLTD |
description |
碩士 === 元智大學 === 工業工程研究所 === 89 === In this study, we use machine vision to defect embedded in homogenously textured surfaces. In order to avoid noise interference in the spatial domain, we employ the Gabor transform method in the spatial-frequency domain to detect local defects. Traditional Gabor-based methods use 2-D Gabor filters for texture analysis. They are computationally intensive and affected by rotation. Given a problem with image size and filter size , the computational complexity of 2-D Gabor filters is .
The proposed method in this study first 1-D ring-projection transformation to compress 2-D images to 1-D signals, and then employs 1-D gabor filters to detect defects. In this way, the computational complexity can be significantly reduced to , and the detection result is invariant to rotation. Both structural textures such as machined surfaces and textile fabrics and stochastic textures such as leather and castings in gray-level and color image are investigated. Experimental results have
shown that the proposed method is effective and efficient for detecting local defects in textured surfaces.
|
author2 |
Du-Ming Tsai |
author_facet |
Du-Ming Tsai Chih-Ping Lin 林志賓 |
author |
Chih-Ping Lin 林志賓 |
spellingShingle |
Chih-Ping Lin 林志賓 Automatic Surface Inspection Using 1-D Gabor Filters |
author_sort |
Chih-Ping Lin |
title |
Automatic Surface Inspection Using 1-D Gabor Filters |
title_short |
Automatic Surface Inspection Using 1-D Gabor Filters |
title_full |
Automatic Surface Inspection Using 1-D Gabor Filters |
title_fullStr |
Automatic Surface Inspection Using 1-D Gabor Filters |
title_full_unstemmed |
Automatic Surface Inspection Using 1-D Gabor Filters |
title_sort |
automatic surface inspection using 1-d gabor filters |
publishDate |
2001 |
url |
http://ndltd.ncl.edu.tw/handle/74743069775361524256 |
work_keys_str_mv |
AT chihpinglin automaticsurfaceinspectionusing1dgaborfilters AT línzhìbīn automaticsurfaceinspectionusing1dgaborfilters AT chihpinglin yīwéijiǎbǎizhuǎnhuànzhībiǎomiànxiácījiǎncè AT línzhìbīn yīwéijiǎbǎizhuǎnhuànzhībiǎomiànxiácījiǎncè |
_version_ |
1716855601843142656 |