基于EM算法的高斯混合型的织物疵点检测研究

为准确检测织物在生产过程产生的疵点，提出了一种基于EM算法的高斯混合模型的算法来实现织物疵点的自动检测。由于织物背景纹理信息对织物疵点检测影响较大，采用均值采样对其进行预处理来消除背景纹理的影响，用高斯混合模型对新得到的图像进行处理。在进行高斯混合模型计算时分为E步骤、M步骤。E步骤初始化参数，计算样本像素的后验概率，M步骤更新高斯混合模型中的各参数。根据计算各像素的后验概率判断各像素点应该属于疵点部分还是非疵点部分。实验结果证明该算法能检测、分割出较多种类的织物疵点，具有较好的有效性和可靠性。

Fabric defect detection based on Gaussian mixture models of EM algorithm

To detect the accurate fabric defects which were generated during the production process, an approach based on Gaussian mixture model of the EM algorithm is presented to achieve automatic detection. Because fabric background texture has greater impact on the fabric defect detection, preprocessing is done by using mean sample to remove its background texture effects, then the Gaussian mixture model is applied to process the new image. Gaussian mixture model calculations are divided into E, M steps. E step is used to initialize the parameters and to calculate the posterior probability of the sample pixel, M steps is used to update the Gaussian mixture model parameters. The defects and background area of the pixel are determined based on their calculated posterior probability. The experimental results show that the proposed algorithm can detect and segment the kinds of fabric defects, and has a good validity and reliability.