基于阴影校正的绝缘子污秽高光谱检测  

Hyperspectral Detection of Insulator Contamination Based on Shadow Correction

作　　者：郭裕钧 刘昆昊 查红原 肖浩然 刘毅杰 吴广宁[1] GUO Yujun;LIU Kunhao;ZHA Hongyuan;XIAO Haoran;LIU Yijie;WU Guangning(School of Electrical Engineering,Southwest Jiaotong University,Chengdu 611756,China)

机构地区：[1]西南交通大学电气工程学院,成都611756

出　　处：《高电压技术》2025年第3期1371-1379,共9页High Voltage Engineering

基　　金：国家自然科学基金(52277165)。

摘　　要：高光谱技术在绝缘子污秽状态的在线大范围检测方面具有较好的应用潜力,然而现场应用环境中由于光照角度和环境遮挡的影响,绝缘子表面难以避免地存在阴影区域,导致高光谱检测误差大,无法准确评估污秽程度和分布。因此,结合阴影识别与分割技术,有效实现了阴影区域的数据校正。通过采集不同形状阴影遮挡下的绝缘子高光谱图像,利用U-Net语义分割网络建立了阴影识别模型。进一步,通过固定视角、改变光照条件获得了多组具有不同阴影分布的绝缘子图像,并将阴影数据校正为相同光照条件下的数据。利用SVM-U Net模型实现了对绝缘子污秽检测高光谱图像中阴影区域的数据校正,由阴影导致的检测平均误差由校正前的15.86%降低至3.21%。Hyperspectral technology has excellent application potentials in on-line large-scale detection of insulator con-tamination status.However,due to the influence of illumination angle and environmental occlusion in the field application environment,there is inevitably a shadow area on the insulator surface,resulting in a large error of hyperspec-tral detection,which makes it impossible to accurately evaluate the degree and distribution of contamination.Therefore,this study combines shadow recognition and segmentation technology to effectively achieve data correction in shadow areas.By collecting hyperspectral images of insulators under shadow occlusion of different shapes,a shadow recognition model is established by using U-Net semantic segmentation network.Further,multiple sets of insulator images with dif-ferent shadow distributions are obtained by fixing the viewing angle and changing the lighting conditions,and the shadow data are corrected to the data under the same lighting conditions.The SVM-U Net model is used to realize the data cor-rection of the shadow area in the hyperspectral image of insulator contamination detection.The average detection error caused by the shadow is reduced from 15.86%to 3.21%.

关 键 词：高光谱成像技术 绝缘子 污秽等级 阴影校正 语义分割 

分 类 号：TM855[电气工程—高电压与绝缘技术]