Vision-Based Object Detection for UAV Solar Panel Inspection Using an Enhanced Defects Dataset
It addresses the problem of efficient solar panel inspection for maintenance, but is incremental as it applies existing methods to a new dataset.
This study evaluated five state-of-the-art object detection models (YOLOv3, Faster R-CNN, RetinaNet, EfficientDet, Swin Transformer) for detecting defects and contaminants on solar panels using a custom dataset, finding trade-offs in accuracy and computational efficiency.
Timely and accurate detection of defects and contaminants in solar panels is critical for maintaining the efficiency and reliability of photovoltaic systems. This study presents a comprehensive evaluation of five state-of-the-art object detection models: YOLOv3, Faster R-CNN, RetinaNet, EfficientDet, and Swin Transformer, for identifying physical and electrical defects as well as surface contaminants such as dust, dirt, and bird droppings on solar panels. A custom dataset, annotated in the COCO format and specifically designed for solar panel defect and contamination detection, was developed alongside a user interface to train and evaluate the models. The performance of each model is assessed and compared based on mean Average Precision (mAP), precision, recall, and inference speed. The results demonstrate the trade-offs between detection accuracy and computational efficiency, highlighting the relative strengths and limitations of each model. These findings provide valuable guidance for selecting appropriate detection approaches in practical solar panel monitoring and maintenance scenarios. The dataset will be publicly available at https://github.com/IsuruMunasinghe98/solar-panel-inspection-dataset.