高斯差分空间的多尺度改进CLBP对带钢表面缺陷的分类

Classification of Strip Surface Defects by Multi-Scale Improved CLBP Based on Gaussian Difference Space

提出一种引入高斯差分空间的改进多尺度完全局部二值模式对带钢表面进行分类,解决由于带钢表面缺陷纹理存在复杂性和多样性,导致对带钢表面缺陷进行分类难度大的问题.首先,根据人类的视觉注意机制,采用高斯差分空间对带钢表面缺陷进行预处理.然后,采用多尺度改进的完全局部二值模式对预处理之后的图片进行特征提取.最后,采用非线性流行学习的方式对特征进行降维,并导入分类器中进行分类.实验结果表明:该方法具有较好的区分性;针对常见的冲孔、污渍、刮边、黑氧化条、结疤等带钢表面缺陷,其最终的分类精度能达到95.7%,优于目前传统的方式.

An improved multi-scale complete local binary model with Gauss difference space is proposed to classify the surface defects of strip steel,which solves the problem that the classification of surface defects is difficult due to the complexity and diversity of surface defect texture.Firstly,according to human visual attention mechanism,Gaussian differential space is used to preprocess the surface defects of strip steel.Then,the multi-scale improved complete local binary model is used to extract the features of the preprocessed image.Finally,the nonlinear popular learning method is used to reduce the dimension of features and import them into the classifier for classification.The experimental results show that the method has good regional characteristics:for common surface defects such as punching,stains,scraping,black oxide strip,scab and other surface defects,the final classification accuracy can reach 95.7%,which is better than the traditional method.