摘要
目的为了准确、快速地对混凝土电镜图像进行自动识别分类,选取合理的特征提取方法以及分类算法,解决预判混凝土耐久性问题.方法利用一种基于局部二值模式(Local Binary Pattern,LBP)的特征提取方法和支持向量机(Support Vector Machine,SVM)分类器的组合模式,对混凝土在电镜下的图像进行特征提取和类别判定,并对SVM分类器选取合理的核函数以及核函数参数.结果采用笔者方法最终图像识别准确率可达79.7%,LBP基本特征算子较好地满足了纹理特征图像的特征提取要求;通过SVM分类器对图像进行分类可以将含有纹理特征的图像成功分类.结论该方法成功有效地识别混凝土的内部多层次结构辅助辨别混凝土的耐久性问题;LBP基本算子和SVM分类算法的组合算法在特征提取与识别分类具有很好的实用性及高效性.
In order to accurately and quickly classify the electron microscopy image,a reasonable feature extraction method and classification algorithm are selected.Using a feature extraction method based on local binary pattern and the combination of support vector machine classifier model,the concrete under the electron microscope image feature extraction and classification and the reasonable kernel function and kernel function parameters are selected for the SVM classifier.Through the method,the final image recognition accuracy rate is 79.7%.It can be seen that the LBP basic feature operator satisfies the feature extraction requirements of texture feature images.The SVM classifier can classify the images and classify the images with texture features successfully.The combination algorithm of LBP basic operator and SVM classification algorithm is verified.The practicability and efficiency of feature extraction and recognition classification.
作者
张丽秀
田甜
王俊海
ZHANG Lixiu;TIAN Tian;WANG Junhai(School of Mechanical Engineering,Shenyang Jianzhu University,Shenyang,China,110168;National-Local Joint Engineering Laboratory of NC Machining Equipment and Technology of High-Grade Stone,Shenyang Jianzhu University,Shenyang,China,110168)
出处
《沈阳建筑大学学报（自然科学版）》
CAS
CSCD
北大核心
2020年第1期148-154,共7页
Journal of Shenyang Jianzhu University：Natural Science
基金
国家自然科学基金项目(51805336)
