水力发电2026，Vol.52Issue(9)：77-85,9.
红外光伏组件缺陷检测技术:融合跨尺度特征的改进YOLO11-EICM方法
Infrared Photovoltaic Module Defect Detection Technology:An Improved YOLO11-EICM Method with Cross-Scale Feature Fusion
陈喜军 1田培根 2刘松民3
作者信息
- 1. 水电水利规划设计总院,北京 100120
- 2. 清华大学,北京 100084
- 3. 国晟世安科技股份有限公司,北京 100093
- 折叠
摘要
Abstract
Infrared detection of photovoltaic module defects is crucial for ensuring the safe operation of photovoltaic power stations and improving maintenance efficiency.However,existing infrared thermal imaging detection methods face significant challenges,such as low contrast,blurred edges,and drastic variations in defect scales,which lead to insufficient edge feature extraction,inefficient cross-scale feature fusion,and high false negative rates for small-scale defects.To address these issues,this paper proposes an improved YOLO11-EICM based infrared detection algorithm for photovoltaic module defects to diagnose module faults.Extensive experiments on an infrared photovoltaic module defect dataset demonstrate that the proposed algorithm achieves a mAP@0.5 of 83.8%and a recall rate of 79.4%,representing improvements of 3.46%and 8.03%respectively over the existing YOLO11n model,while reducing the parameter count by 8.91%to just 2.35 million.关键词
红外热成像/光伏组件缺陷检测/YOLO11/边缘信息增强/多尺度特征融合/深度学习Key words
infrared thermal imaging/photovoltaic module defect detection/YOLO11/edge information enhancement/multi-scale feature fusion/deep learning分类
信息技术与安全科学引用本文复制引用
陈喜军,田培根,刘松民..红外光伏组件缺陷检测技术:融合跨尺度特征的改进YOLO11-EICM方法[J].水力发电,2026,52(9):77-85,9.
