Nitrogen is an important element affecting the growth of winterwheat. The real-time application of nitrogen fertilizer based on the demand of winterwheat is one of the key problems to be solved in modern agricultural precision fertilization. Unmanned Aerial Vehicles (UAV) remote sensing technology has the advantages of high resolution, high timeliness and low cost in the monitoring of winterwheat growth, which provides an important data source for solving the problem ofwinter wheat fertilizer demand monitoring. Therefore, studying the multi-spectral image data of UAV and constructing its relationship model with winter wheat yield and fertilization is very important for precision fertilization research. This study carried out field trials with four different kinds of nitrogen levels in a typical production area of winter wheat in Huantai, Shandong. The multispectral images of winter wheat at the returning green stage were collected from experimental area with different nitrogen fertilization levels using Sequoia multispectral sensor equipped with UAV. Meanwhile, winter wheat canopySoil and Plant Analyzer Development (SPAD) and yield were measured. Six vegetation index such as NDVI, SAVI and MCARI2 were obtained after calculation, and established UAV multispectral images vegetation indexes and the winter wheat canopy SPAD of linear function, quadratic polynomial function, logarithm function, exponential function and power function, to screen out the sensitivity index of winter wheat canopy reflecting different nitrogen levels. Further, according to the relationships of different nitrogen fertilization levels with sensitive vegetation indexes and winter wheat yield, a variable nitrogen fertilization model based on vegetation indexes was constructed and applied to simultaneous images. The results are as follows: (1) SPAD could reflect the nitrogen fertilization level and growth of winter wheat, and the canopy reflectance of winter wheat with different nitrogen fertilization levels varied greatly. (2)The structural vegetationindex and SPAD fit better than other types of index. and the optimal vegetation index of the estimation model established based on SPAD was MCARI2 (R2=0.790, RMSE=0.22), which was considered as the sensitive vegetation index of nitrogen fertilizer. (3) Based on the yield-nitrogen fertilizer model and yield-sensitive vegetation index model, the variable rate fertilization model of nitrogen fertilizer was Nr=10 707.63×MCARI22-5 992.36×MCARI2+715.27. Based on the model, a variable nitrogen fertilization map for winter wheat was produced in the experimental area, which was highly consistent with actual fertilization. In this study, the model and method of nitrogen fertilization for winter wheat based on UAV multispectral data was proposed, which provides areference for the precise fertilization of winter wheat.