To establish a rapid, efficient and low-cost method of applying Fourier-Transform Near-Infrared Reflectance Spectroscopy (FT-NIRS) to detecting amino acids in soybean, we selected 167 representative soybean flour samples from a large original population and utilized FT-NIRS to predict the contents (which were analyzed by HPLC) of eighteen amino acid species. According to the data we obtained, aspartate (R2CV = 0.85), glutamate (R2CV = 0.86), serine (R2CV = 0.82), glycine (R2CV = 0.89), tyrosine (R2CV = 0.83), phenylalanine (R2CV = 0.78), isoleucine (R2CV = 0.86), ryptophane (R2CV = 0.81) and the total amino acids content (R2CV = 0.82) can be detected by FT-NIRS accurately; the detected models of threonine, arginine, alanine, valine, leucine, and cystine are valuable as references and can be used to estimate the relative contents; the content prediction of histidine, lysine, proline and methionine is inaccurate. This study has further demonstrated the stability and feasibility of using FT-NIRS to detect the major amino acid components in soybean.