基于灰度共生矩阵和稳健马氏距离的织物横档类疵点检测  

Fabric's Filling Bar Defect Detection Based on Grey-Level Co-occurrence Matrix and Robust Mahalanobis Distance

作　　者：张向东[1,2] 黄秀宝[1] 

机构地区：[1]东华大学纺织学院,上海201620 [2]大连工业大学纺织轻工学院,辽宁大连116034

出　　处：《东华大学学报（自然科学版）》2009年第6期691-698,共8页Journal of Donghua University(Natural Science)

摘　　要：在分析横档类疵点纹理特点的基础上,提取织物纹理图像的灰度共生矩阵单特征值——对比度.利用最小中值平方估计的快速算法,获得正常织物纹理训练样本的稳健马氏距离,并应用契比晓夫不等式确定在一定置信度条件下判断待检织物为疵点的马氏距离的阈值.对8种不同纹理结构、织物密度和纱线线密度的织物进行了横档类疵点的检测,在90%置信度下,可检出90%以上的横档类疵点,误检率为3.28%,检测效果较好.Based on the analysis of textural characteristic of filling bar defect of woven fabric, the feature value of grey-level co-occurrence matrix was extracted from the fabric images. Then the robust Mahalanobis distances of training samples with normal texture were calculated by using the fast algorithm for the estimation of least median squares. The threshold value of Mahalanobis distance was set up for detecting the defect with Chebyshev inequality under certain confidence level. Eight fabrics with different fabric texture, fabric count and yarn count were detected and the results were satisfied. More than 90% of filling bar defect can be found out, and the misdetection rate was only 3.28% under 90% confidence level.

关 键 词：横档类疵点 灰度共生矩阵 最小中值平方估计 稳健马氏距离 

分 类 号：TD101.8[矿业工程—矿山地质测量]