Spectral matching has strict requirements to stability and accuracy of the spectral curve, while the spectral curve in hyperspectral image data is influenced by many factors. An automatic search flat field strategy is designed based on the flat field method. An improved preprocessing method is proposed for hyperspectral images on the basis of dark-object subtraction and empirical method commonly used in the other research areas. Thus automatic reflectance inversion based on hyperspectral image data is realized. Experiments using three different images show that the method of this paper is more accurate and robust than the traditional methods, and can effectively improve the effect of spectral matching.