The Comparison of Remote Sensing Estimation Models for Fractional Vegetation Cover Based on UAV Hyperspectral Image

The UAV hyperspectral data on a banana forest land in Yicang Village, Minzhong Township, Zhongshan City, Guangdong Province is used to explore the accuracy and applicability of the spectral unmixing model for estimating vegetation coverage. The effects of three classical models (pixel binary model, Carlson model and Baret model) and three spectral unmixing models (LMM model, PPNMM model and NCM model) in estimating vegetation coverage are compared. The experimental results show that the pixel binary model overestimates the vegetation coverage; the Carlson model underestimates the vegetation coverage; the Baret model overestimates the vegetation coverage in the low vegetation coverage area and underestimates the vegetation coverage in the high vegetation co-verage area; the LMM model has good estimation effect in areas with high vegetation coverage; the PPNMM model overestimates slightly in low vegetation coverage; and the NCM model has the best estimation effect.