机构地区: 暨南大学信息技术研究所
出 处: 《计算机与数字工程》 2009年第12期147-150,共4页
摘 要: 在基于磁瓦表面缺陷图像直方图、纹理、投影和形状的特征提取的基础上,提出了一种用LVQ神经网络进行缺陷分类的方法,对现场采集到的6种主要缺陷类型进行了试验。试验结果表明,基于LVQ神经网络的分类器训练与分类的时间短,多缺陷种类分类时准确率高。 The LVQ neural network classification method was introdued based on feature extraction of arc segments ceramic magnet for histogram, texture, projection, shape. Testing by 6 main defect types collected from online was made. The results indicated that the surface defects classification based on LVQ neural network spent little time for training and classifying, and its accuracy was higher.
领 域: [自动化与计算机技术] [自动化与计算机技术]