Estimating soil moisture conveniently and exactly is a hot issues in water resource monitoring among agriculture and forestry. Estimating soil moisture based on vegetation index has been recognized and applied widely. 8 vegetation indexes were figured out based on the hyper-spectral data measured by portable spectrometer. The higher correlation indexes among 8 vegetation indexes and surface vegetation temperature were selected by Gray Relative Analysis method (GRA). Then, these selected indexes were analyzed using Multiple Linear Regression to establish soil moisture estimation model based on multiple vegetation indexes, and the model accuracy was evaluated. The accuracy evaluation indicated that the fitting was satisfied and the significance was 0.000 (P<0.001). High correlation was turned out between estimated and measured soil moisture with R2 reached 0.636 1 and RMSE 2.149 9. This method introduced multiple vegetation indexes into soil water content estimating over micro scale by non-contact measuring method using portable spectrometer. The exact estimation could be an appropriate replacement for remote sensing inversion and direct measurement. The model could estimate soil moisture quickly and accurately, and provide theory and technology reference for water resource management in agriculture and forestry.