Improved YOLO Series Models for Automated Photovoltaic Panel Defect Detection

Abstract

During long-term outdoor operation, photovoltaic (PV) panels are highly susceptible to various defects, such as hot spots and diode short circuits, due to environmental erosion and inherent material characteristics. Existing traditional detection methods exhibit significant limitations when handling defects with irregular shapes, microscopic sizes, or complex backgrounds, failing to meet the industry's urgent demand for high-precision and high-efficiency automated inspection. To address these challenges, this study proposes an improved algorithm based on the YOLO series models. To enhance the feature extraction capability for defects with irregular geometric shapes, we replaced the standard Bottleneck structure in the C2f module with a Bottleneck_DCN module based on Deformable Convolution v2 (DCNv2), thereby endowing the network with a more flexible receptive field. Furthermore, to mitigate the high miss rate of microscopic defects, we introduced the HyperC2Net architecture, which significantly improves the model's sensitivity to small targets by strengthening multi-scale feature fusion mechanisms. Experimental results demonstrate that the improved algorithm outperforms the original YOLOv8 series models across multiple key metrics, achieving a 1.34% higher Precision than YOLOv8. This study not only provides an efficient and reliable technical solution for PV panel defect detection, filling the gap in accuracy and adaptability of existing methods, but also offers important theoretical foundations and practical references for promoting intelligent operation and maintenance and technological progress in the PV industry, demonstrating significant engineering application value and social benefits.