摘要
为实现织物疵点自动检测,本文应用Mallat算法对织物图像进行小波分解,并根据织物组织结构特点,对纬向子图和经向子图分别提取能量、熵、极差、方差和逆差矩5个特征值,同时以平纹坯布为实验对象,对白杠、断纬、破洞、反丝、杂纤维、掉扣、飞花、色污等多种常见疵点进行特征值提取实验,实验结果表明,织物疵点使织物图像在经纬方向产生不规则纹理,通过提取特征值可得出织物疵点图像在不同特征值曲线处均有明显变化。该研究对判定疵点的存在、分析疵点的走向与"形状"具有参考价值。
In order to realize automatic fabric defect detection, Mallat algorithm has been used to wavelet decomposition of a fabric image. According to fabric texture＇s structure characteristics, impl five ement elgenvalues, namely the energy, entropy, extreme difference, variance and inverse difference, were extracted from weft sub-image and warp sub-image respectively. In this paper, the tabby cloth was chosen as experimental object, and the eigenvalue extraction experiments were carried on various common defects, such as white streak, weft-lacking, broken hole, anti-silk, miscellaneous fiber, buckle-off, loom fly, unclean color and so on, and the eigenvalue changes caused by different ＂strip-shape＂ defects and ＂area-shape＂ defects were analyzed, and the correlativity between eigenvalue changes and defect trend. The results showed that the method has the inference value to determine the existence of defects, defect analysis of the direction and ＂shape＂ has the reference value.
出处
《青岛大学学报（工程技术版）》
CAS
2013年第2期53-59,共7页
Journal of Qingdao University(Engineering & Technology Edition)
基金
青岛市科技局基础研究项目(12-1-4-2-(9)-jch)
关键词
织物疵点
计算机视觉
小波变换
图像处理
特征提取
fabric defect
computer vision
wavelet transform
image processing
feature extraction
