Electroluminescence (EL) imaging has emerged as a critical diagnostic technique for identifying defects in solar panels during production, installation, and operation. This non-destructive method effectively detects microcracks, broken finger lines, and other anomalies that can lead to significant power losses if left unaddressed. However, the manual interpretation of EL images is often hindered by complex fault patterns, heterogeneous backgrounds, and the need for expert-level domain knowledge, making it a labor-intensive process. To overcome these limitations, this study proposes an advanced automated visual inspection system that leverages state-of-the-art object detection algorithms—YOLOv8 and YOLOv9—for efficient and accurate classification of defects in solar cells. These algorithms are trained on annotated EL image datasets to identify and categorize various types of faults with high precision and speed. The integration of deep learning strategies into the inspection workflow not only accelerates defect detection but also enhances reliability and scalability in real-world solar panel monitoring systems. This approach ultimately contributes to reduced power outages, improved maintenance efficiency, and prolonged module lifespan.

Keywords: Electroluminescence Imaging, Solar Panel Defects, YOLOv8, YOLOv9, Automated Visual Inspection, Deep Learning, Microcrack Detection, Solar Cell Monitoring, Image Classification, Fault Detection.