基于二维Otsu算法的织物疵点检测

Defect detection of fabric based on 2D Otsu algorithm

为了准确检测织物的疵点,提出一种基于二维Otsu算法的织物疵点检测方法。首先采用均值滤波对采集的织物疵点进行预处理,减少高斯噪声对图像质量影响的同时,也有效地抑制了织物背景纹理信息对织物疵点检测的影响;然后对处理后的图像采用二维Otsu算法进行阈值分割;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,去除毛疵点和孤立点等。实验结果表明:对比其他检测方法,综合主观视觉效果和客观峰值信噪比(PSNR)值,该方法在织物疵点检测中既能有效保留图像的边缘信息,也不损伤图像的细节质量,检测效果较好,在疵点检测方面具有一定的实用价值。

We presented a new method based on 2D Otsu algorithm to accutately detect the fabrice defect. Firstly,the fabric image sample was preprocessed using the mean filter to reduce the influence of Gaussian noise and background texture on fabric defects; then the processed defective images were segmentated by 2D Otsu algorithm; finally,the contour of the segmented images were smoothed and the wool defect and isolated points were removaled using morphological operation. Based on comprehensive subjective visual effect and objective PSNR values,the results showed that the presented method not only retain the defect images edge information but also no damage to the details of the image quality. The detection effect was great and had certain practical value in filed of fabric defect detection.