摘要
根据疵点形态特征上的差异而采用不同检测方案的思想,在图像的空间域中,采用改进的标准差方法检测油污、破洞等没有方向性的疵点;在图像频率域中,结合二阶统计分析,用纬向的矩形窗口分割图像小波分解的垂直分量并检测缺纬、双纬等纬向疵点;用经向的矩形窗口分割图像小波分解的水平分量并检测断经、双经等经向疵点。为确定窗口分割的最佳宽度,研究提出一种基于灰度差分和窗口分割迭代算法。实验表明:方法实现了油污、破洞、断经、缺纬4类疵点的准确检测与分类。
The idea of selecting different detecting schemes for fabric defects of different shapes was proposed in this paper. In the space domain of the inputted image, the improved standard deviation method was adopted to detect those no-directional defects such as oil dirty and break holes. In the frequency domain of the inputted image, combining the second-order statistics analysis, the vertical component of wavelet decomposition was segmented by a series of weft rectangular windows, and those weft defects such as mispick, double weft could be detected from these windows; the level component of wavelet decomposition was segmented by a series of warp rectangular windows, and those warp defects such as broken warp, double warps could be detected. In order to determine the optimal width of the segmentation windows, a new iterative algorithm based-on gray difference was proposed. Experiments showed that the proposed method can accurately detect and simultaneously classify the defects of those four types： oil dirty, break holes, broken warp and mispick.
出处
《丝绸》
CAS
北大核心
2009年第7期24-27,共4页
Journal of Silk
关键词
织物疵点检测
小波分析
灰度差分
窗口分割
Fabric defect detection
Wavelet analysis
Gray-scale difference
Window segmentation
作者简介
王学文（1979-），男，硕士研究生，研究方向为图像处理技术在纺织中的应用。
通讯作者：邓中民，教授，hzcad@163．com。
