摘要
针对盘形悬式瓷绝缘子串红外图像边缘缺失导致铁帽和伞盘误分割的问题,提出一种基于边缘检测的绝缘子串图像分割方法。首先对红外图像进行灰度化、中值滤波和去噪等图像预处理,然后通过比较多种边缘检测算法,采用PSNR方法确定了多尺度形态学梯度算法是对绝缘子串边缘提取中准确性和完整性最高的方法。接着对二值化后的图像确定最大连通区域,对其进行边缘分割得到绝缘子串区域。最后结合边缘检测确定面积阈值,采取区域提取方法对其进行铁帽和伞盘区域分割,得到铁帽和伞盘区域。通过大量样本训练,正确率达到90%以上,在实际应用中取得十分良好的效果,对提高绝缘子红外智能检测准确率有重要指导意义。
To address the problem of the improper segmentation of the iron cap and the porcelain disk surface due to the missing edge in the infrared image of the porcelain pin type insulator string,an insulator image segmentation method based on the edge detection is proposed. Firstly,infrared image is processed by graying,median filtering and denoising. Secondly,by comparing multiple edge detection algorithms,the PSNR method determines that multi-scale morphological gradient algorithm is the best way to extract insulator string edge,both accurately and completely. Thirdly,the binarized image helps to gain the maximum communication area,and the edge cut apart produces obtain the insulator string area. Finally,by using the area threshold obtained from the edge detection,a specific algorithm was applied to segment the iron cap and the disk surface area to obtain the iron cap and the disk area. Through a large number of sample training,the correct accuracy rate reaches up to more than 90%,which is effective in practical applications and significant to improve the accuracy of infrared detection of insulators.
作者
刘洋
陆倚鹏
高嵩
毕晓甜
尹骏刚
朱向前
姚建刚
LIU Yang;LU Yipeng;GAO Song;BI Xiaotian;YIN Jungang;ZHU Xiangqian;YAO Jiangang(Jiangsu Electric Power Company,Nanjing 210024,China;College of Electrical and Information Engineering,Hunan University,Changsha 410082,China;Hunan HDHL Electric&Information Technology Company Limited,Changsha 410285,China)
出处
《电瓷避雷器》
CAS
北大核心
2020年第1期198-203,共6页
Insulators and Surge Arresters
基金
国家电网公司总部科技项目资助.
关键词
绝缘子串
边缘检测
PSNR
样本训练
分割
insulator string
edge detection
PSNR
sample training
segmentation
