Monitoring of Winter Wheat Biomass Using UAV Hyperspectral Texture Features

Biomass is an important indicator to evaluate vegetation life activities and hyperspectral imagery from unmanned aerial vehicle (UAV) supplied with abundant texture features shows a great potential to estimate crop biomass. In this paper, principal component analysis (PCA) was used to select the principal component bands from UAV hyperspectral image. Eight texture features from the principal component bands were extracted by Gray Level Cooccurrence Matrix method, and the sensitive texture features were finally selected to construct the biomass estimation model. The results show that: (1) Texture features mean, ent, sm, hom, con, dis of the first principal component (pcal) and the mean of the third principal component (pca3) were significantly correlated with the biomass. (2) The biomass model by multiple texture features (R-2 = 0.654, RMSE = 0.808 (10(3) kg/hm(2))) demonstrated better fitting effect than that by single texture feature (R-2 = 0.534, RMSE = 0.960 (10(3) kg/hm(2))). The biomass estimation model based on the texture features of multiple principal components had a good fitting effect. Therefore, texture features of the UAV platform can accurately predict the winter wheat biomass.