基于小波特征和SVM的织物疵点识别

针对常见织物疵点具有方向性,利用传统空间域特征识别方法不能有效定位局部疵点区域且当样本较少时分类率低的问题,为有效定位疵点且提高分类率,提出了水平和垂直方向上小波域特征,利用能有效解决小样本分类问题的支持向量机进行分类识别;并对利用图像灰度共生矩阵特征及小波域特征的分类结果进行了比较。仿真实验结果表明,所选特征不仅能对织物疵点区域进行水平和垂直方向上的定位,而且得到了较高的正确分类率。

Fabric Defects Identification Based on Wavelet Transform and SVM

In view of that common fabric defects have directivity,it used the problems that traditional spatial characteristics can not effectively locate defect region and the accuracyis always unsatisfactory when samples are less,and in order to effectively locate faults and improve the classification rate,it proposed a method of wavelet domain features on the horizontal and vertical direction,used support vector machine to effectively solve limited sample classification problem,compared classification results between the traditional features of GLCM(gray level co-occurrence matrix)and wavelet domain features.Experiment results showed that using wavelet features can locate the areas of fabric defects on the horizontal and vertical position and also can receive a higher classification accurate.