Classification of desert steppe species based on unmanned aerial vehicle hyperspectral remote sensing and continuum removal vegetation indices

Grassland species composition is a key indicator for assessing grassland ecological health. In order to obtain the distribution of species in the desert steppe of Gegen tara, Inner Mongolia, China, we used a new hyperspectral imaging system based on an unmanned aerial vehicle to acquire images. Through spectral transformation, the spectral difference of species was increased. The classification features of desert steppe species were constructed by spectral transformation vegetation indices. The maximum interclass variance was used to compare the reflectance vegetation index, the first order transformation differential vegetation index, the logarithm transformation vegetation index and the continuum removal transformation vegetation index, to select the classification features and to calculate the classification threshold. The classification decision tree model constructed by continuum removal transformation normalized differential vegetation index and continuum removal transformation difference vegetation index has the best effect on desert steppe species classification, and the overall classification accuracy and Kappa coefficient are 87% and 0.8. The classification and identification of keystone species Stipa breviflora Griseb., Artemisia frigida Willd., and Salsola collina Pall. in the desert steppe were realized by this new method. Our study will have a positive effect on the understanding of the evolutionary law of grassland ecosystems, and provide quantitative indicators for grassland ecological management.

Automated summary

This study used a new hyperspectral imaging system based on an unmanned aerial vehicle to obtain the distribution of species in the desert steppe of Gegen Tara, Inner Mongolia, China. Through spectral transformation and classification features construction, accurate classification and identification of desert steppe species were achieved, providing important quantitative indicators for grassland ecological management.