基于Mean Shift滤波的织物疵点检测方法

Fabric defect detection based on Mean Shift filtering

为了实现自动织物疵点检测,提出了一种基于Mean Shift滤波的织物疵点检测方法。该算法首先求取织物样本的熵图像,反映原始图像信息的变化程度,然后对熵图像进行Mean Shift平滑滤波,达到去除噪声和增强织物疵点部分的目的,从而利于疵点的分割,最后对滤波后的图像进行阈值分割,得到二值化检测结果。对6种纹理织物进行处理,共检测出18种疵点,92.5%的疵点可以被成功检测并定位。另外,实验中还将Mean Shift滤波与Gabor滤波的检测结果进行了比较,结果表明Mean Shift滤波对某些类型的疵点的检测效果更为理想。

In order to realize the automatic fabric defect detection,a fabric defect detection method based on Mean Shift filtering is proposed in this paper. Firstly,the entropy image of a fabric sample is calculated to reflect the degree of information change in the original image. Then Mean Shift filtering is performed on the entropy image in order to remove the noise and strengthen the defects,which is benefit for the defect segmentation. Finally,adaptive threshold is used to identify the filtered image as defective ones and the binary image is obtained. 6 kinds of texture were detected and 18 kinds of defect were found in this paper,and 92. 5% of defects can be successfully detected and located. In addition,the detection results of Mean Shift filtering are compared with the results of Gabor filtering. The result indicates that Mean Shift filtering brings about more ideal results in several types of defects.