基于改进迭代匹配滤波的织物疵点检测

Fabric defect detection based on improved iterative match filter algorithm

为准确有效地检测织物的疵点,给出一种基于改进迭代匹配滤波的织物疵点检测算法.该算法首先采用多种滤波方法对样本疵点图像进行对比预处理,以去除样本图像的高斯噪声,减少织物背景纹理信息对织物疵点检测的影响,并根据去噪图像的客观峰值信噪比与主观视觉效果等确定均值采样为最优的预处理结果.然后利用改进的迭代匹配滤波器对去噪后的织物疵点进行滤波并分割出疵点.最后对分割后的图像进行形态学运算后处理、平滑图像轮廓和去除毛疵点和孤立点等.实验结果表明,该方法在织物疵点检测中能有效保留图像的边缘细节信息,可有效地检测较多种类织物的瑕疵,而且检测精度较高.

In order to detect fabric defects effectively and accurately,a new method based on improved iterative match filter algorithm is presented. Firstly, the fabric image samples are comparatively preprocessed by multiple filtering methods to reduce the Gaussian noise in the sample image and the influence of background texture on fabric defects. Meanwhile, the mean sampling algorithm is determined as the optimal preprocessed result according to objective peak signal to noise ratio and subjective visual effect of the denoised image. Then the processed images are filtered and segmented by improved iterative matching filter. Finally, the contour of the segmented images are smoothed,the defect and isolated points are removed using morphological operation. Experimental results show that the presented method can not only effectively preserve the edge details of the image,but also detect a variety of the fabrics defects. In addition, it has high accuracy and runs fast.