AI-powered denoising boosts clarity in outdoor PV imaging study

Researchers from the Technical University of Denmark and the University of New South Wales, Australia, have developed a deep learning-based method to enhance electroluminescence imaging for outdoor photovoltaic module inspections. The method has used a simplified ResNet architecture, SimpleResNet, which has outperformed conventional denoising techniques in preserving detail and improving image clarity. A preprocessing pipeline was also designed, combining distortion correction, denoising, and sharpness enhancement to address challenges under outdoor imaging conditions. The researchers say that the method is optimized in terms of speed, which makes it feasible for high-throughput inspection in utility-scale PV power plants. Real-world electroluminescence images of different PV module types and lighting conditions were used for validating the method, proving to be robust and field applicable.