基于稀疏分析的织物缺陷检测方法

A Fabric Defect Detection Method Based on Sparse Analysis

针对具有纹理背景分布的织物表面缺陷检测效率低、受主观影响大等问题,根据图像在冗余字典下可以稀疏表示的特点,建立缺陷图像的稀疏表示模型,提出基于稀疏分析的缺陷检测算法.该算法将缺陷图像分割为缺陷前景部分和纹理背景部分,分割后的前景图像去掉纹理背景,则缺陷区域完全被凸显,从而实现缺陷的提取.在多种不同缺陷类型的织物表面图像进行实验,实验结果表明,本文算法对有结构性纹理背景的织物表面缺陷检测正确率达到96.75%,验证了算法的有效性.

In order to solve the problems of low efficiency and strong subjectivity in fabric surface defect detection with texture background distribution,a sparse representation model of defect images is established based on the fact that images can be sparsely represented in redundant dictionaries.A defect detection algorithm based on sparse analysis(SADD)is proposed.In this algorithm,the defect ima ge is divided into defect foreground part and texture background part.The texture background is removed from the segmented foreground image,and the defect area is completely highlighted so as to realize the defect extraction.The effectiveness of the algorithm is verified by experiments in various fabric surface images of different defect types.The experimental results show that the accuracy rate of the algorithm for fa bric surface defect detection with structural texture background is 96.75%.