一种方向性纹理织物疵点的检测方法

A Detection Method of Directional Texture Fabric Defects

首先,利用Hough获取织物的纹理主方向及其正交方向,由Gabor滤波器沿着这两个方向分别进行滤波,取模值图像为输出;然后,应用最大熵对两个输出模值图像进行二值化分割,融合这两个分割后的图像并进行形态学处理和去除孤立点;最后,得到疵点图像检测结果.实验选取5种织物疵点进行验证,结果表明:该方法针对不同纹理方向的织物都有良好的检测效果,且滤波器数量少,无需事先学习.

Firstly,using the Hough transform to obtain the main direction and the orthogonal direction of texture,then taking these direction of texture as the direction of Gabor filter,the model image was taken as an output characteristics;and then using maximum entropy to segment these amplitude images,fusing the subimages and morphology processing, removing the outlier detection;finally,the fabric defect image was achieved.The experiment select five types of fabric de-fects,the experimental results shows that the method has good detection effect for different texture of fabric,few filters were needed and without prior learning.