基于希尔伯特边际谱互相关分析的织物疵点检测  

Fabric Defect Detection Based on Cross-Correlation of the Hilbert Marginal Spectrum

作　　者：王帅军[1] 袁晔[1] 李丹华[1] 

机构地区：[1]北京服装学院信息工程学院,北京100029

出　　处：《北京服装学院学报（自然科学版）》2015年第4期15-20,26,共7页Journal of Beijing Institute of Fashion Technology：Natural Science Edition

摘　　要：提出了一种基于希尔伯特黄变换和互相关分析相结合的织物检测疵点方法.首先,选取正常织物作为模板并对待检测图像进行等尺寸分割,得到子图像,其大小与模板尺寸保持一致;然后分别对模板图像和每个子图像做希尔伯特黄变换,得到希尔伯边际谱.最后,求取模板图像的边际谱和每个子图像边际谱的互相关,选取合适的互相关值作为疵点识别的依据.仿真试验表明,该方法对断经、断纬、缺经、粗经、错经、错纬、扣痕、破洞等疵点具有良好的检测效果.A new method is proposed in this paper to defect fabric defects based on cross-correlation of the Hilbert marginal spectrum.Firstly,a piece of normal fabric is chosen as the template whose size is the same as that of the subimage to be defected.Secondly,the Hilbert-Huang transform（HHT） is performed on the template and each subimage to be defected to obtain their respective marginal spectrum.Finally,the cross-correlation is computed for the Hilbert marginal spectrum corresponding to the template image and the subimage respectively.An appropriate cross-correlation values is selected as the threshold of the defect identification.Simulation results have shown that this method is valid for various defects,such as broken weft,broken warp,missing warp,coarse end,mixed and pick,buckle mark,broken hole.

关 键 词：希尔伯特黄变换 疵点检测 互相关 边际谱 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]