摘要
针对高速铁路接触网绝缘子故障检测问题,提出一种基于深度学习EAST模型与Hu不变矩的图像检测方法。利用EAST模型生成检测图像的掩膜图,与图像进行掩膜操作得到绝缘子区域,实现定位功能;利用二值化边缘检测提取绝缘子每片轮廓;利用Hu不变矩逐片进行轮廓相似度对比,通过对比结果判断绝缘子是否存在故障。实验结果表明,本文方法可在不同场景下准确定位绝缘子区域,通过逐片对比避免了图像一致性差的问题,为绝缘子故障检测提供一种思路。
To solve problem of the fault detection of high-speed railway catenary insulators,an image detection method based on deep learning EAST model and Hu invariant moment was proposed in this paper.The EAST model was used to generate the mask image of the detected image,and the image was masked to obtain the insulator area to realize the positioning function.The binary edge detection was used to extract the contour of each insulator.The Hu invariant moment was used to compare the contour similarity.By comparing the results,the insulator fault can be determined.The experimental results show that the method proposed in this paper can accurately locate the insulator region under different scenes.The comparison of the images one by one avoids the problem of poor image consistency,which provides a way for insulator fault detection.
作者
张子健
马吉恩
李旭峰
方攸同
ZHANG Zijian;MA Jien;LI Xufeng;FANG Youtong(College of Electrical Engineering,Zhejiang University,Hangzhou 310027,China;China Academy of West Region Development,Zhejiang University,Hangzhou 310012,China;School of Mechanical&Automotive Engineering,Zhejiang University of Science&Technology,Hangzhou 310023,China)
出处
《铁道学报》
EI
CAS
CSCD
北大核心
2021年第2期71-77,共7页
Journal of the China Railway Society
基金
国家自然科学基金(51577166,51637009)。
关键词
深度学习
HU不变矩
绝缘子
故障检测
deep learning
Hu invariant moment
insulator
fault detection
