基于图像的换流站用避雷器泄漏电流识别
何红太1，崔家瑞2，纪建才2，胡广大2，张波3，刘广一3
（1．北京国网富达科技发展有限责任公司，北京市丰台区100070；2．北京科技大学自动化学院，北京市海淀区100083；3．中国电力科学研究院，北京市海淀区100192）
关键词：避雷器泄漏电流仪表；图像矫正；图像预处理；改进Hough变换；读数识别
中图分类号：TP391 文献标志码：A
摘要：针对电力系统中的换流站用避雷器泄漏电流仪表的特点，提出了一种基于图像的换流站用避雷器泄漏电流的识别方法。首先，采用基于透视变换的方法对采集到的避雷器泄漏电流仪表图像进行畸变矫正。然后，通过图像预处理进行图像增强获得相对比较清晰的表盘信息。其次，利用减影法获取指针的位置。再次，采用基于Canny(坎尼)算子边缘检测和改进的Hough（霍夫）变换检测出避雷器泄漏电流表中的指针直线，完成了换流站避雷器泄漏电流的识别。最后，采用现场拍摄的型号为3EX5 050的西门子避雷器泄漏电流表对本文所提算法进行了验证。结果表明该算法克服了传统Hough变换的许多缺点，能够快速准确的识别出避雷器泄露电流大小。该方法避免了人工读取效率低和精度差的问题，而且只需要很少的改动便可应用于换流站中其他指针式仪表的智能识别。
A Recognition Method of Leakage Current of Arrester
in Converter Station Based on Image
HE Hong-tai1, CUI Jia-rui2, JIJian-cai2, HU Guang-da2, ZHANG Bo3,LIU Guang-yi3
(1. State Grid Fuda science and Technology Development Limited LiabilityCompany, Fengtai District, Beijing 100070, China; 2. School of Automation andElectrical Engineering, University of Science and Technology Haidian District,Beijing 100083, China; 3.China Electric Power Research Institute, HaidianDistrict, Beijing 100192, China)
Key words：leakage current of arrester; imagerectification; image processing; Hough transform; recognition of reading
Abstract：According to the characteristics ofthe converter station lightning arrester leakage current meter in the powersystem, a kind of recognition method of the converter station lightningarrester leakage current meter was put forward based on image. Firstly, adistortion correction was made on the collectedarrester leakage current meter image based on the method of perspective transform.Then, relatively clear dial information was obtained according to the imageprocessing for image enhancement. Secondly, the subtraction method was used toget a more precise pointer position. Thirdly, the Canny operator edge detectionand the improved Hough transform were used to detect the pointer line of the converterlightning arrester leakage current meter,and the identification of converterstation lightning arrester leakage current was completed. Finally, Siemensarrester leakage current meter of the 3EX5 050 were conducted to validate the proposedalgorithms. The results shows that the algorithm overcomes many disadvantagesof the traditional Hough transform and has the ability to identify sizearrester leakage current fast and accuratly . This method can not only avoidthe low artificial reading efficiency and the problem of poor accuracy, but alsobe applied to intelligent recognition of other pointer type instruments inconverter station with very few changes.