用于电力设备异常诊断的图像配准及融合方法  

Image Registration and Fusion Method for AnomalyDiagnosis of Power Equipment

作　　者：周喜红 席亚宾 李中宝 ZHOU Xi-hong;XI Ya-bin;LI Zhong-bao(Guangdong Yuedian Daya Bay Integrated Energy Co.,Ltd,Huizhou 516000,China;China Nuclear Industry 23 Construction Co.,Ltd,Beijing 101300,China)

机构地区：[1]广东粤电大亚湾综合能源有限公司,广东惠州516000 [2]中国核工业二三建设有限公司,北京101300

出　　处：《电工电气》2024年第11期67-76,共10页Electrotechnics Electric

摘　　要：近年来图像融合方法在电力设备热异常的诊断中所占比重逐渐增加,但是涉及到图像配准和融合统一考虑的方法很少。提出了一种最大迭代关联图像配准及区域特性判别的图像融合方法,用于辅助热异常的诊断。该方法通过构建约束函数计算源图像配准迭代次数,隶属度函数定义源图像的区域特性,已知区域特性的子图像根据电力设备热异常所重视的特征优先选择融合策略,以最大程度保留源图像中的纹理特征和热辐射特征。在自建的电力设备数据集上与其他方法对比显示,所提方法在保证源图像配准精度的前提下,还突出了红外图像的热辐射特征和可见光图像的纹理特征,能够满足电力设备热异常诊断的需要。In recent years,the proportion of image fusion methods in the diagnosis of thermal anomalies of power equipment has gradually increased,but the methods involving unified consideration of image registration and fusion are rare.Therefore,this paper proposes an image fusion method based on maximum iterative correlation image registration and regional feature discrimination,which is used to assist thermal anomaly diagnosis.This method calculates the number of source image registration iterations by constructing a constraint function,and the membership function defines the regional characteristics of the source image.The sub-images with known regional characteristics preferentially select the fusion strategy according to the characteristics that the thermal anomaly of the power equipment attaches importance to,so as to retain the texture features and thermal radiation features in the source image to the greatest extent.Compared with other methods on the self-built power equipment dataset,the proposed method not only ensures the registration accuracy of the source image,but also highlights the thermal radiation characteristics of the infrared image and the texture characteristics of the visible image,which can meet the needs of thermal anomaly diagnosis of power equipment.

关 键 词：图像融合 图像配准 电力设备 热异常诊断 约束函数 隶属度函数 热辐射 

分 类 号：TM711[电气工程—电力系统及自动化] TP391[自动化与计算机技术—计算机应用技术]