AI and the World's Hard Problems · Life sciences · Nature (computational biology) · 2026-10-25
Multi-scale vision transformer model with UAV-based hyperspectral imagery for early identification of pine wilt disease infected trees
Pine wilt disease spreads rapidly and causes high tree mortality, so control at the green attack and early stages is critical. The study developed a multi-scale vision transformer for UAV-based hyperspectral imagery, aimed at identifying infected trees in these stages, where standard RGB images show few visible cues. The model achieved 87.40% and 88.21% accuracy for the green attack and early stages, an overall accuracy of 87.90% and a Kappa coefficient of 83.25%, outperforming traditional machine learning and other deep learning models with only 2.45 M parameters.