摘要
针对纺织行业织物疵点检测自动化的需求,提出了基于二维经验模态分解(BEMD)的织物疵点分割方法的改进。在BEMD算法中,使用基于Delaunay三角化(DT)的三次样条分段插值替代基于径向基函数(RBF)的全局插值,以提高计算效率和分解有效性。在BEMD的分解结果中,选择第二个和第三个内蕴模式函数(IMF)进行融合后进行分割以提高疵点分割结果的完整性。实验中以多幅典型的疵点织物为样本,对比了不同插值方法和分割对象的检测误差率(DER),结果显示改进后的疵点分割方法具有更好的计算效率和鲁棒性。
For the demand of automatic fabric defect detection in textile manufactures, improvements of fabric defect segmentation based on Bidimensional Empirical Mode Decomposition(BEMD)are provided. In the BEMD, the Radial Basis Function(RBF)based global interpolation is replaced by the Delaunay Triangulation(DT)based piecewise interpolation with the cubic spline to obtain higher computation efficiency and decomposition effectiveness. In the BEMD results, the second Intrinsic Mode Function(IMF)and the third IMF are fused to improve the integrity of defect segmentation. In the experiments on multiple typical sample images, the Detection Error Rates(DER)of using different interpolation techniques and segmentation objects are compared; and the experiment results demonstrate that the improved defect segmentation scheme owns a better calculation efficiency and robustness.
作者
厉征鑫
刘建立
周建
高卫东
LI Zhengxin;LIU Jianli;ZHOU Jian;GAO Weidong(School of Textile and Clothing, Jiangnan University, Wuxi, Jiangsu 214122, China)
出处
《计算机工程与应用》
CSCD
北大核心
2016年第24期217-222,共6页
Computer Engineering and Applications
基金
江苏省2011年度普通高校研究生科研创新计划项目(No.CXZZ11_0472)
国家自然科学青年基金(No.61203364)
高等学校博士学科点专项科研基金(No.20120093130001)
关键词
织物疵点分割
二维经验模态分解
内蕴模式函数
径向基函数
DELAUNAY三角化
fabric defect segmentation
bidimensional empirical mode decomposition
intrinsic mode function
radial basis function
Delaunay triangulation