The development of modern technology has made hyperspectral sensors much smaller in size and lighter in weight, which can be carried by unmanned aerial vehicles (UAVs). A new type of imaging spectroradiometer based on UAV was used to acquire the hyperspectal images of cotton field, which were used to establish the regression model aiming to predict the SPAD value of cotton leaf and make its distribution map. The results showed that in the wavelength range of 720～850nm, the reflectance had positive correlation with SPAD value. Many spectral indexes based on the hyperspectal images were significantly correlated to the SPAD value of cotton leaf on P<0.01 level. The absolute correlation coefficients of four indexes, including DR526, DR578, SDy and Db were all above 0.8. DR526, DR578, SDy and Db were used to establish the simple regression model of SPAD respectively. All the spectral indexes whose absolute correlation coefficients with SPAD value were above 0.7 were chosen to establish the multiple regression inversion model of SPAD using multiple stepwise regression(MSR) method and partial least squares regression(PLSR) method. According to the accuracy test, both SPAD-MSR model and SPAD-PLSR model had high accuracy to predict the SPAD value of cotton leaf. The six inversion models of SPAD were used to make the distribution map of cotton leaf SPAD value. The map using SPAD-PLSR model had the best result which was the closest to real SPAD distribution. Thus this research provides a new technology to supervise the growth information of cotton and other crops.