Please wait a minute...
J4  2010, Vol. 44 Issue (3): 448-452    DOI: 10.3785/j.issn.1008973X.2010.03.006
    
Surface defect detection based on multivariate image analysis
 WANG Xuan-Yin, LIANG Dong-Tai
Download:   PDF(0KB) HTML
Export: BibTeX | EndNote (RIS)      

Abstract  

 A new approach based on multivariate image analysis (MIA) was proposed to improve the accuracy and reliability of surface defect detection for industrial product quality control. Multivariate test images are constructed by stacking both the standard defectfree images and the test image. MIA technique uses multiway principal component analysis (PCA) to obtain the principal component scores of the multivariate test images. As the feature space for defect detection, the Qstatistic image is derived from the residuals which are left after the extraction of the first principal component and the noise. In the Qstatistic image, surface defects can be efficiently detected by using an appropriate threshold. Experimental results showed the effectiveness of the proposed method. Compared with the gray histogram based method, the proposed method has less sensitivity to the inhomogeneous of illumination, and has more robustness and reliability of defect detection with lower pseudo reject rate.



Published: 20 March 2012
CLC:  TP391.41  
Cite this article:

WANG Xuan-Yin, LIANG Dong-Tai. Surface defect detection based on multivariate image analysis. J4, 2010, 44(3): 448-452.

URL:

http://www.zjujournals.com/eng/10.3785/j.issn.1008973X.2010.03.006     OR     http://www.zjujournals.com/eng/Y2010/V44/I3/448


基于多元图像分析的表面缺陷检测算法

为了提高工业产品质量控制中表面缺陷检测的准确性和可靠性,提出一种基于多元图像分析(MIA)的表面缺陷检测算法.采用合格样本图像和被测图像堆叠构成多元被测图像,利用基于主成分分析(PCA)的多元图像处理方法,获得多元被测图像的主分量表示,将去掉第一主分量和噪音后的Q统计图像作为缺陷特征的检测空间,利用阈值处理检测缺陷.金属罐内壁的表面缺陷检测实验结果表明,与现有基于直方图阈值处理等检测方法相比,受照明不均匀影响小,提高了图像检测系统的准确性和鲁棒性,使缺陷的误检率大大降低.

[1] MALAMAS E, PETRAKIS G M, ZERVAKIS M, et al. A survey on industrial vision system, applications and tools [J]. Image and Vision Computing, 2003, 21(2): 171188.
[2] NIKHIL R P, SANKAR K P. A review on image segmentation techniques [J]. Pattern Recognition, 1993, 26(9): 12771294.
[3] 张润楚. 多元统计分析 [M]. 北京:科学出版社,2006.
[4] GELADI P, GRAHN H. Multivariate Image Analysis [M].Chichester U K: Wiley, 1996.
[5] ESBENSEN KH, GELADI P. Strategy of multivariate image analysis [J]. Chemometrics and Intelligent Laboratory Systems, 1989, 7(12): 6786.
[6] BHARATI MH, MACGREGOR JF. Multivariate image analysis for realtime process monitoring and control [J]. Industrial & Engineering Chemistry Research, 1998, 37(12): 47154724.
[7] BHARATI MH, MACGREGOR JF. Texture analysis of images using Principal Component Analysis [C]// SPIE/Photonics Conference on Process Imaging for Automatic Control. Boston, MA: SPIE,2000: 2737.
[8] BHARATI M H, MACGREGOR J F, TROPPER W. Softwood lumber grading through online multivariate image analysis techniques [J]. Industrial and Engineering Chemistry Research, 2003, 42(21): 53455353.
[9]  PRATSMONTALBAN J M, FERRER A. Integration of colour and textural information in multivariate image analysis: defect detection and classification issues [J]. Journal of Chemometrics, 2007, 21(12): 1023.
[10] GELADI P, ISAKSSON H, LINDQVIST L, et al. Principal component analysis of multivariate images [J]. Chemometrics and Intelligent Laboratory Systems, 1989, 5(3): 209220.
[11] NOMIKOS P, MACGREGOR J F. Multivariate SPC charts for monitoring batch processes [J]. Technometrics, 1995, 37(1): 4159.
[12] GONZALEZ R C, WOODS R E. Digital Image Processing [M], 2th ed. New Jersey, USA: Prentice Hall, 2002.

[1] SUN Zhi-Hai, KONG Mo-Ceng, SHU Shan-An. Improved algorithm of subtractive clustering for object location  in video sequences[J]. J4, 2010, 44(3): 458-462.
[2] FAN Xiang, XIA Shun-ren. Feature based automatic stitching of microscopic images[J]. J4, 2009, 43(7): 1182-1186.
[3] ZHANG Dong-Mei, LIU Li-Gang. Angle-filtering based smoothing algorithm for planar graphs[J]. J4, 2009, 43(6): 1042-1046.
[4] CHEN Cheng, PENG Huo-Ting, XIAO Dun. Video foreground segmentation with camera movement[J]. J4, 2009, 43(6): 975-977.