Hot Spot Localization Method for Photovoltaic Panels Based on Deep Learning Techniques

The infrared hot spot detection task for photovoltaic panels is an important mission to ensure the safe and stable operation of photovoltaic power generation systems. However, the low-resolution characteristics of infrared images and the relatively small size of hot spots compared to the overall image make it challenging for current object detection algorithms to address this issue. To solve this problem, we propose a Dyhead multi-attention mechanism detection head improved based on the DCNv3 method and integrate it into YOLOv8 to form the PVPHSDNet model, which effectively overcomes these challenges. Through comparative experiments in this study, PVPHSDNet significantly outperforms the original YOLOv8 algorithm and achieves a favorable balance between accuracy and speed. The proposed model provides a reliable tool for drone-based inspection and detection of hot spots in photovoltaic panels.