全部 标题 作者
关键词 摘要

OALib Journal期刊
ISSN: 2333-9721
费用:99美元

查看量下载量

相关文章

更多...

基于特征提取的缺陷图像分类方法

Keywords: 缺陷图像,特征提取,缺陷分类,BP神经网络

Full-Text   Cite this paper   Add to My Lib

Abstract:

针对缺陷图像表面复杂多变、特征不宜提取的特点,提出了一种归一化转动惯量特征和不变矩特征相结合的时域分析方法来构建缺陷图像的统计特征量,同时增加缺陷矩形框区域内压缩度、距离极值比和线度特征量作为缺陷分类的依据;提出了在缺陷频谱图像内提取特征量的频域分析方法,并将矩形框区域内所有像素点灰度平均值和灰度方差值作为缺陷分类的另一重要依据;同时将BP神经网络应用于缺陷图像的自动分类中,构建了系统的缺陷分类器,并对现场采集的常见6种缺陷类型进行了实验.结果表明,该特征提取方法在很大程度上提高了特征的分类有效性;该BP分类器识别率较高,现场整体识别率达到90%以上,在一定程度上解决了缺陷图像分类难的问题.

References

[1]  杨志晓,冯冬青,陈铁军,等.人工神经网络技术在无损检测中的应用[J].无损检测,2002,24(6):244-250.YANG Zhi-xiao,FENG Dong-qing,CHEN Tie-jun,et al.Application of articifical neural networks to nonesturctive[J].Nondestructive Testing,2002,24(6):244-250.(in Chinese)
[2]  WU Gui-fang.Application of a newfeature extraction and optimization method to surface defect recognition of cold rolled strips[J].Journal of University of Science and Technology Beijing,2007,14(5):437-442.
[3]  CHIOU Yih-chih.The feature extraction and analysis of flawdetection and classification in BGAgold-plating areas[J].ExpertSystems with Applications,2007,85(8):1-9.
[4]  ANIRBAN Mukherjee.Image-based classification of defects in frontal surface of fluted ingot[J].Measurement,2006,8(7):687-698.
[5]  杨小冈,付光远,缪栋.基于图像NMI特征的目标识别方法[J].计算机工程,2002,28(6):6-9.YANG Xiao-gang,FU Guang-yuan,MIAO Dong.A new approach to target recognition based on image NMI feature[J].Computer Engineering,2002,28(6):6-9.(in Chinese)
[6]  陈斌,万江文,吴银锋,等.神经网络和证据理论融合的管道泄漏诊断方法[J].北京邮电大学学报,2009,32(1):9-13.CHEN Bin,WANG Jiang-wen,WU Yin-feng,et al.A pipeline leakage diagnosis for fusing neural network and evidencetheory[J].Journal of Beijing University of Posts and Telecommunicaations,2009,32(1):9-13.(in Chinese)
[7]  KAMARTHI S V,PITTNER S.Accelerating neural network training using weight extrapolations[J].Science on NeuralNetworks,2000,12(6):1285-1299.
[8]  MARTIN F,MOLLER S.A scaled conjugate gradient algorithm for fast supervised learning[J].Neural Networks,2000,6(3):525-533.
[9]  YAN Wei-ping.The study of gas sensor array signal processing with improved BP algorithm[J].Sensors and Actuators B,2000,66(7):283-285.
[10]  JIANG Ming-hu.A fast learning algorithm for time-delay neural networks[J].Information Sciences,2002,24(3):27-39.

Full-Text

Contact Us

service@oalib.com

QQ:3279437679

WhatsApp +8615387084133