Year: 2025
Type: article
Abstract: Regular cleaning of the photovoltaic (PV) panel is crucial for maintaining optimal photovoltaic power generation efficiency. However, manual cleaning methods for PV panels are often inadequate and costly, underscoring the need for the introduction of PV cleaning robots. A significant challenge in the development of these robots is the recognition of PV panel poses. We propose the YOLOv8n-Photovoltaic-Pose (YOLOv8n-PP) method, a lightweight pose recognition algorithm, to address this issue. It is specifically designed for PV panel cleaning robots. Built upon the YOLOv8 framework, YOLOv8n-PP incorporates the Mobile-ViT visual module, which is both lightweight and mobile-friendly. This integration helps mitigate the effects of varying target poses from the robot’s mobile perspective. Additionally, we introduce the LMPDIoU boundary box regression loss to enhance the precision of PV panel recognition. Furthermore, we have developed a diverse and comprehensive dataset for PV panel poses to improve the model’s generalization capabilities. Our method shows improvements in both precision and recall, providing an effective solution for PV pose recognition.
Authors: Jidong Luo, Guoyi Wang, Yanjiao Lei, Dong Wang, Hongzhou Zhang
Institution: Tarim University
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $3,290
OpenAlex ID: W4411785995