Flavor is an important factor affecting meat quality and consumption. It is mainly converted from some flavor precursors through complex chemical reactions. There is a significant correlation between mutton flavor and oleic acid （C18：1） content， and stearic acid （C18：0） content is positively correlated with the unpleasant flavor of mutton. In this paper， the visible/near infrared hyperspectral （400-1 000 nm） was used to quickly detect the content of characteristic fatty acids in flavor precursors in Tan Mutton. First select the spectral image after the mask as the region of interest， adopt six methods for spectral preprocessing， and compare and analyze to select the best preprocessing method. Combining successive projection algorithm （SPA）， uninformative variable elimination（UVE） and variable combination cluster analysis （VCPA） extracts characteristic wavelengths and establishes a regression prediction model. Finally， the fatty acid content is quantitatively inverted on the sample image through the established optimal prediction model. The results show that oleic acid has the best modeling effect with MSC-UVE-MLR， and its RC=0.9099， RP=0.8759， and stearic acid has the best modeling effect with SNV-SPA-MLR， whose RC=0.8691， RP=0.8446. The use of spectral reflectance values combined with optimal model parameters achieves a good visual representation of the fatty acid content on the hyperspectral image of lamb samples， enabling quick and non-destructive testing of lamb quality in a more intuitive and clear way.