基于类别共生矩阵的纹理疵点检测方法

根据有规则纹理的特点,提出了基于类别的共生矩阵来描述纹理特征,从而很好地将正常纹理与疵点区分开.分析了传统的灰度共生矩阵在计算纹理特征时计算量大,且分辨能力差的缺点.为了克服灰度共生矩阵在计算量和分辨能力上的缺点,定义了类别共生矩阵.在类别共生矩阵的算法中,首先学习纹理的一些基本特征以确定类别共生矩阵的一些关键参数,如纹理的概率密度分布、纹理的主方向和周期,以及分类准则等重要参数,然后计算类别共生矩阵并提取白疵点增强、黑疵点增强和一致度等三个特征,最后采用异常点检测的方法即可很好地区分正常纹理和疵点.实验证明,该方法比已有的灰度共生矩阵计算量小,并具有更突出的分辨纹理和疵点的能力.

Textural defect detection based on label co-occurrence matrix

In accordance with the characters of the well-regulated texture, a method of the label co-occurrence matrix (LCM) was proposed to depict the textural features and discriminate the defects from the normal texture. The traditional gray-level co-occurrence matrix (GLCM) was computationally intensive and discriminatively insufficient. The proposed method of LCM was ameliorated mainly from computational load and discriminability. In LCM algorithm, some basic textural features were acquired initially to fix some key parameters of LCM, such as the probability density distribution of the texture, dominant orientation and periodicity of the texture, classification regulations, and etc. LCMs were computed and some simple features such as white defect emphasis, black defect emphasis and consistency were extracted from those LCMs. The outlier detection for those features was applied to detect textural defects. Experiments proved that the method was characterized by less computation and better discriminability in comparison with the existing GLCM.