融合可见光与红外图像的光伏阵列缺陷检测  

DEFECT DETECTION OF PHOTOVOLTAIC ARRAYS BY FUSING VISIBLE AND INFRARED IMAGES

作　　者：白晓静[1] 徐佳伟 皮宇啸 张文彪[1] 洪烽 李佩哲 Bai Xiaojing;Xu Jiawei;Pi Yuxiao;Zhang Wenbiao;Hong Feng;Li Peizhe(School of Control and Computer Engineering,North China Electric Power University,Beijing 102206,China)

机构地区：[1]华北电力大学控制与计算机工程学院,北京102206

出　　处：《太阳能学报》2025年第4期313-321,共9页Acta Energiae Solaris Sinica

基　　金：中央高校基本科研业务费专项资金(2021MS016)。

摘　　要：为保证光伏发电稳定、高效和安全运行,需及时检测光伏阵列的运行状况并发现存在的缺陷。提出融合可见光与红外图像的光伏阵列缺陷检测方法,采用CenterNet进行可见光图像光伏组件中太阳电池检测,采用U-Net对红外图像高温区域进行分割,提出区域匹配模块对可见光与红外图像进行匹配,提出关键点(PoI)聚集模块和二次分类器实现关键点处特征向量的聚集以及太阳电池缺陷分类,最后结合可见光图像异物遮挡及红外图像温度异常识别缺陷太阳电池位置及类型。选择不同网络进行测试,提出的算法在较为轻量的CenterNet和U-Net网络上太阳电池检测的AP50-95值达到84.4%,异常温度区域分割的IoU达到89.7%,且单张检测时间约为38 ms,能以较快的速度完成异常太阳电池的检测。In order to ensure the efficient and safe operation of photovoltaic power generation,it is necessary to detect the operating status of photovoltaic arrays.A defect detection method of photovoltaic array based on the fusion of visible light and infrared images is proposed.CenterNet and U-Net are employed to detect solar cells in visible images and segment the high temperature region of infrared images,respectively.A region matching module is proposed to match visible and infrared images.Point of Interest(PoI)gathering module and secondary classifier are proposed to achieve the aggregation of feature vectors at key points and the classification of solar cells defects.Finally,defective solar cells are identified according to the visible and infrared image recognition results.Different networks are selected for testing.The value of AP50-95 of solar cells detected by the proposed algorithm on the relatively light CenterNet and U-Net networks reaches 84.4%,the IoU of abnormal temperature region segmentation reaches 89.7%,and the time cost of a single image is about 38ms.The results show that the proposed method can complete the detection of abnormal solar cells at a faster speed.

关 键 词：太阳电池 光伏阵列 目标检测 图像分割 红外图像 可见光图像 

分 类 号：TK513.5[动力工程及工程热物理—热能工程]