Fine Crop Classification Based on UAV Hyperspectral Images and Random Forest

The classification of unmanned aerial vehicle hyperspectral images is great significance in agricultural monitoring. This paper studied a fine method for crops based on feature transform combined with random forest (RF). Aiming at the problem large number spectra and amount calculation, three methods dimensionality reduction, minimum noise fraction (MNF), independent component analysis (ICA), principal (PCA), were studied. Then, RF was used to finely classify variety images. results showed: (1) MNF–RF combination best ideal this study. accuracies sample set Longkou Honghu areas 97.18% 80.43%, respectively; compared original image, accuracy improved by 6.43% 8.81%, respectively. (2) For study, overall two regions positively correlated points. (3) image after less affected points than image. MNF curve varied but smoothest least points, followed PCA ICA curves. did not exceed 0.50% 3.25%, respectively, fluctuation research can provide reference UAV-borne