Energies, Год журнала: 2024, Номер 18(1), С. 119 - 119

Опубликована: Дек. 31, 2024

Amidst the dual challenges of energy shortages and global warming, photovoltaic (PV) power generation has emerged as a critical technology due to its efficient utilization solar energy. Rooftops, underutilized spaces, are ideal locations for installing panels, avoiding need additional land. However, accurate generalized segmentation large-scale PV panel images remains technical challenge, primarily varying image resolutions, large scales, significant imbalance between foreground background categories. To address these challenges, this paper proposes novel model based on Res2Net architecture, an enhanced version classic ResNet optimized multi-scale feature extraction. The integrates Spatial Feature Reconstruction aggregation modules, enabling effective extraction data features precise reconstruction spatial features. These improvements particularly designed handle small proportion panels in images, effectively distinguishing target from redundant ones improving recognition accuracy. Comparative experiments conducted publicly available rooftop dataset demonstrate that proposed method achieves superior performance compared mainstream techniques, showcasing effectiveness segmentation.

Язык: Английский