摘要
针对电力设备图像识别问题,提出基于多特征联合稀疏表示的新方法。分别采用Krwatchouk矩、LBP描述子以及PCA特征矢量描述电力设备图像的几何形状、局部纹理以及像素分布特性。因此,三种特征具有良好互补性。在分类阶段,采用联合稀疏表示对三类特征进行表征,利用它们的内在关联提高稀疏表示的精度。最终,根据三类特征的总体重构误差判定测试样本的类别、采用绝缘子、变压器和断路器三种典型电力设备图像对提出方法进行性能测试,结果表明了方法的有效性。
This paper proposes an electric equipment image recognition method via joint sparse representation of multiple features.The Krwatchouk moments, LBP describtors, and PCA feature vector are used to describe the geometrical shape, local texture, and intensity discretion of the electric equipment image. Hence, the three features could complement each other. During the classifcaiton stage, the joint sparse representation is employed to represent the three features toghtther. By exploiting their inner correlations, the precision of sparse representations can be enhances. Finally, the label of the electric equipment image is decided according to the total reconstruction error of all the three features. Experiments are conducted on the images of insulators, power transformers, and breakers to evaluate the performance of the proposed method. The results show its effectiveness.
作者
乔林
孙宝华
徐立波
刚毅凝
代东旭
QIAO Lin;SUN Bao-hua;XU Li-bo;GANG Yi-ning;DAI Dong-xu(State Grid Liaoning Electric Power Co.,Ltd.,Information and Communication Branch,Shenyang 110006 China)
出处
《自动化技术与应用》
2020年第11期120-123,共4页
Techniques of Automation and Applications
