基于支持向量机的胶片缺陷检测算法

Algorithm of Film Defects Detection by Support Vector Machines

提出了一种基于支持向量机的胶片缺陷检测算法,该算法把胶片中的非缺陷区域和缺陷区域分别看作两种不同的纹理模式,应用机器学习理论的年轻分支———支持向量机对两类不同的样本采样学习,然后进行分类判断。实验结果表明,这种算法能够较好地实现胶片缺陷的检测分类,有着深入研究的价值。文中使用了两种不同的方法进行图像的特征提取,它们是主元分析法和离散余弦变换法,结果显示,不同的特征提取方法对训练分类的结果会产生一定影响。

This paper presents an algorithm about film defects detection by Support Vector Machines that is the new branch of machine learning,in which the defective area and non-defective area are treated as two different textures and are sampled respectively to be learned for classification.It is shown that this algorithm is worth researching because it works well in defects detection.Two different feature extraction methods,PCA and DCT,are applied to show their effects on the result.