Abstract: According to the spectrum difference between fabric texture and defect, a new method for defect detection is presented. On the basis of the fabric texture and defect classification, a spectrum image is derived from FFT. The frequency-domain filter is designed to remove normal texture information, based on the image reconstruction, fabric normal texture and defects are isolated by image segmentation. Experimental results show this method is valid and feasibile.

Key words: statistical texture, structural texture, frequency-domain filter, defect detection

摘要： 根据织物纹理和疵点频谱的不同分布，提出织物疵点检测的新方法。在织物纹理和疵点分类的基础上，通过傅里叶变换获得频谱图。设计频域滤波器抑制正常纹理频谱信息，通过重构对灰度图像进行分割，实现疵点与正常织物纹理的分离。实验结果表明该方法是有效的。

关键词: 统计性纹理, 结构性纹理, 频域滤波器, 疵点检测

CLC Number: 

• TP182