Research on Infrared Hot Spot Detection Algorithm of Photovoltaic Module Based on Faster Regionbased Convolutional Neural Network

With the application and development of photovoltaic (PV) power generation, defects in its operation can significantly reduce the efficiency of power generation, so infrared thermography is crucial for hot spot defect detection in PV modules. In this paper, an improved Faster Region-based Convolutional Neural Network (Faster R-CNN) algorithm is proposed for automatic hot spot detection in aerial infrared images. First, the anchor frame size configuration is optimized based on the clustering algorithm to address the challenges of small hot spot target size and complex background. Subsequently, the ResNet-50 backbone network is used instead of the original network to enhance the hot spot feature extraction capability. Finally, it is trained and tested on the self-constructed thermal image dataset, and the experimental results show that the detection accuracy of this method is improved compared with the traditional method, which provides technical support for the intelligent operation and maintenance of PV on the UAV edge computing platform.