Volume 36: Intelligent Energy Solutions for Resilient Urban Systems

Identification and localization of Photovoltaic Defect from UAV Thermal Infrared Data Based on Photogrammetry and Deep Learning Bo Wang, Mengmeng Wang, Xiaodan Shi, Qi Chen, Jian Yang, Zhengjia Zhang



This study presents an efficient framework for locating and classifying faulty Photovoltaic (PV) panels from Unmanned Aerial Vehicle (UAV) thermal infrared images. First, aerial triangulation based on photogrammetry is used to obtain thermal infrared images of PV panels with coordinate information, then, individual PV panels are segmented based on High-Resolution Network (HRNetV2-W32), finally, the panels are fed into residual net (ResNet-50) to classify the fault types. Results showed that the panel segmentation accuracy reaches 98.54%, the classification accuracy reaches 88.74%, and the coordinate error is better than 0.033m.

Keywords Photovoltaic, Deep Learning, Semantic Segmentation, Defect Classification, Thermal Infrared

Copyright ©
Energy Proceedings