基于奇异值分解的双算法织物缺陷检测

针对难以有效地同时检测洞形缺陷和线形缺陷问题，提出一种基于奇异值分解的双算法织物缺陷检测方法。该方法首先对图像进行奇异值分解，通过对原图与特征值图进行布尔差集运算消除背景纹理并保留缺陷区域；然后采用均值滤波、直方图均值化及方差阈值滤波消除纹理及噪声点的干扰；接着通过形态学处理确定缺陷位置；最后采用面积阈值和方差阈值的方式获取线形缺陷和洞形缺陷。实验结果表明：该方法不仅能够有效地检测洞形缺陷，而且在检测线形缺陷上也有很好的表现，并在准确率上明显高于传统算法，证明了本文算法的有效性和多用途性。

Dual-algorithm for fabric defect detection based on singular value decomposition

Aiming at the problem that hole and line defects are difficult to be detected simultaneously, a double-algorithm fabric defect detection method based on singular value decomposition was proposed. The image was decomposed by singular value first, and then the background texture was eliminated and the defect area was preserved by Boolean difference set operation between the original image and the eigenvalue image. Following that the mean filtering, histogram average and variance threshold filtering were used to eliminate the interference of texture and noise points and the defect position was determined by morphological processing. The linear defects and hole defects were eventually obtained by using area threshold and variance threshold. Experimental results show that the proposed method not only can effectively detect hole defects, but also has a good performance in detecting linear defects, and the accuracy is significantly higher than the traditional algorithm, which proves the effectiveness and versatility of the proposed method.