Abstract: 【Objective】 To explore the normalized difference vegetation index, an estimation model for maize yield,this study has provided a theoretical basis for the multispectral prediction of maize yield by field UAV. 【Methods】 Xinjiang eighteen spring-sown maize was taken as the research object, a multispectral image of the flowering period was obtained, and the actual spectral reflectance of the experimental area was obtained after radiometric correction, atmospheric correction, establishment of mask, extraction of NDVI map, and calculation of vegetation coverage, so as to obtain the actual value of normalized vegetation index. In addition to that, the NDVI value was fitted to the measured yield in the field. 【Results】 The power functions Y = 23,411.46-10,997.99 / X (R² = 0.488,6), polynomial functions Y = 39,003.00-117,963.03X + 103,130.25X₂ (R² = 0.562,0), using partial least squares regression. The result was a linear function Y = 24,458.22X-9,620.55 (R² = 0.722,2). 【Conclusion】 The results show that it has a certain correlation with maize yield, and the linear function model of NDVI value and yield during the critical period of maize growth.