Abstract: In view of the fact that GPS receiver is sensitive to the water content of surrounding vegetation when receiving L-band signal, the normalized difference vegetation index (NDVI) of the station is retrieved by using the change of GPS reflected signal. In this paper, the normalized microwave reflection index (NMRI) calculated from the continuous observation data of two GPS reference stations in recent five years is used to construct a univariate linear model for NDVI inversion. The correlation coefficients R of the inversion results are 0.626 53 and 0.625 73 respectively, and the root mean square error root mean square error (RMSE) are 0.051 29 and 0.055 08 respectively. The correlation coefficients of the inversion results are increased by 2% and 6% respectively by using BP neural network model, which indicates that the regional NDVI inversion results of GPS-interferometric reflectometry (GPS-IR) have high reliability. This study provides a theoretical support for obtaining accurate position, real-time continuous and high-resolution NDVI.