基于核模糊聚类的机织物组织自动识别

本文基于数字图像处理技术，采用无监督的模式识别方法，实现机织物组织的自动识别。首先，采用错切变换对倾斜纱线进行校正，并按图像的经、纬向灰度投影曲线实现织物组织点的定位。然后，根据灰度共生矩阵对组织点图像进行纹理特征的提取。为了减少数据冗余量，对组织点特征进行主分量分析，提取最有意义的子特征。最后，采用无监督的核模糊C均值聚类对组织点进行分类识别。织物的测试结果表明，所采用的算法能够实现简单的机织物组织的准确识别，并输出对应的组织图。

Automatic identifying of fabric weave patterns based on kernel fuzzy clustering

This paper employ unsupervised pattern recognition to automatically identify the weave patterns of the fabric based on digital image processing technology. Firstly, it is using shear to correct of the tilt yarn, and gray projection curves of warp and weft to locate crossed points. Then texture features of each crossed-area image are extracted based on gray level co-occurrence matrix(GLCM). In order to reduce the amount of data redundancy, principal component analysis must be implemented to extract the most significant child feature. Finally, the kernel fuzzy C-means clustering is applied to classify the crossed points. The experimental results show that the algorithm can identify the basic fabric weave patterns correctly and output the chart of the identified weave pattern.