一种基于混合纹理特征的板材表面缺陷检测方法  

A Surface Defect Detection Method for Sheet Metal Based on Mixed Texture Features

作　　者：吴志轩 黄少楠 黄远民[1] 杨伟杭 刘欣达 Wu Zhixuan;Huang Shaonan;Huang Yuanmin;Yang Weihang;Liu Xinda(Foshan Polytechnic,Foshan,China)

机构地区：[1]佛山职业技术学院,广东佛山

出　　处：《科学技术创新》2024年第16期30-33,共4页Scientific and Technological Innovation

基　　金：广东省科技创新战略专项资金项目(板材表面缺陷自动检测系统;课题项目编号:pdjh2023a0994)。

摘　　要：目前,板材表面缺陷检测利用传统人工检测很难满足客户要求。为此,本文提出了一种基于混合纹理特征的表面缺陷检测算法,能准确、鲁棒地检测出板材表面图像中是否有缺陷。采用基于融合后的混合特征向量,对纹理特征项目进行有效组合,提取图像纹理采用了灰度共生矩阵方法。应用BP人工神经网络对样本集进行训练和检测。该方法能准确地对板材表面缺陷进行检测,平均检测成功率达97.2%,该检测方法满足企业要求,提高了检测效率,降低了成本。该检测技术,值得运用和推广。At present,it is difficult to meet customer requirements for surface defect detection of sheet metal using traditional manual inspection.Therefore,this article proposes a surface defect detection algorithm based on mixed texture features,which can accurately and robustly detect whether there are defects in the surface image of the sheet metal.The texture feature items were effectively combined using a mixed feature vector based on fusion,and the image texture was extracted using the gray level co-occurrence matrix method.Train and detect the sample set using BP artificial neural network.This method can accurately detect surface defects on the board,with an average detection success rate of 97.2%.This detection method meets the requirements of enterprises,improves detection efficiency,and reduces costs.This detection technology is worth applying and promoting.

关 键 词：混合纹理特征 灰度共生矩阵 BP人工神经网络 

分 类 号：S781.5[农业科学—木材科学与技术] TP391.4[农业科学—林学]