In order to predict crop yields efficiently and accurately, winter wheat was taken as the research object, a UAV remote sensing platform was used, and a hyperspectral camera was carried to obtain UAV images of each growth stage to estimate crop yields. In order to accurately predict the yield, according to the characteristics of hyperspectral with more spectral information and the unique red edge area, nine vegetation indices and five red edge parameters were selected. The correlation between vegetation indices and red edge parameters and yield was analyzed. Five vegetation indices and two red edge parameters were selected for constructing yield estimation models, and then three yield estimation models with different growth stages were constructed: single-parameter linear regression model, model based on vegetation indices using partial least squares regression method, model based on vegetation indices combined with red edge parameters and using partial least squares regression method, and using different models to estimate winter wheat yield. The results showed that most of the vegetation indices and red edge parameters of the four growing stages were very significantly correlated with yield. Single-parameter linear regression models constructed at the jointing, flagging, flowering and filling stages, with the best performing parameters being REP, Dr/Drmin, GNDVI and GNDVI. The partial least squares regression method was used to improve the accuracy of yield estimation. At the same time, the model constructed with the vegetation indices combined with the red edge parameters as the factor improved the yield estimation effect (better than the yield model constructed with the vegetation indices as the factor). The research result provided a reference for UAV hyperspectral to estimate crop yield in agriculture.