摘要
针对直升机巡检输电线路所拍摄的彩色图像,提出一种新的航拍玻璃绝缘子图像分割方法.该方法首先利用彩色图像中玻璃绝缘子的颜色呈现聚类的特性,通过K均值聚类算法将绝缘子从复杂的背景图像中大致分割出来;然后利用绝缘子的形状特性,采用基于主成分分析的连通区域判决方法定位出绝缘子的具体位置.实验结果表明了此方法的有效性,而且与现有的一些玻璃绝缘子图像分割算法相比,该方法受不同光线和复杂背景影响较小,提高了绝缘子分割的准确性和鲁棒性.
This paper proposes a novel segmentation approach for glass insulators in color images acquired byhelicopter patrol inspection system. The method firstly uses the color clustering feature of glass insulators in col-or images to separate them from complex background images by K mean clustering algorithm roughly,and thenlocates their specific positions according to the approach of judging connecting regions based on principal com-ponent analysis. Experiments show that our method is effective. Furthermore, compared with some availablemethods about defect detection of glass insulators,our method is affected less by different light conditions andcomplex backgrounds and has more robustness and reliability for the segmentation of glass insulators.
作者
胡建平
李玲
谢琪
张道畅
Hu Jianping;Li Ling;Xie Qi;and Zhang Daochang(Science College,Northeast Electric Power University,Jilin Jilin 132012;School of Mathematical Science,Jilin Univer-sity,Changchun Jilin 130012)
出处
《东北电力大学学报》
2018年第2期87-92,共6页
Journal of Northeast Electric Power University
基金
国家自然科学基金(61672149)
吉林省科技发展计划基金(20170520052JH)
吉林省教育厅十三五科学技术研究基金(吉教科合字[2016]第97号)
关键词
玻璃绝缘子
K均值聚类
主成分分析
图像分割
Glass insulator
K mean clustering
Principal component analysis
Image segmentation
