To determine fatty acid content in camellia oil non-destructively and precisely, the method was proposed to determinfatty acid content in camellia oil by near infrared spectroscopy technology. 156 kinds of commercial camellia oil were used as the samples camellia oil fatty acid composition and content was got by using the gas chromatograph, and the near infrared spectrum information collected by using the spectrometer. The correlation between the original spectrum, SG smooth spectral, two derivative transform spectrum and the fatty acid content in camellia oil was respectively analyzed. The optimizing quantitative model was obtained to determine the fatty acid content in camellia oil through the comparison between the full wavelengths and the significant wavelengths using partial least squares regression methods. . The results indicated that main fatty acids of camellia oil were palmitic acid, oleic acid and linoleic acid, which ranged from 4.428% to 10.931%, from 78.036% to 84.621%, from 7.013% to 9.863%, respectively. The spectrum scanning test for camellia oil samples showed that the characteristic variation of spectrum among camellia oil samples was located in 8 600–8 200, 7 300–6 900, 6 000–5 500, 4 800–4 500 and 4 500–4 000 cm–1, respectively. A positive correlation between the content of palmitic acid in camellia oil and the absorbance of R, SG spectrum was observed, and there was a negative correlation between the content of oleic acid and linoleic acid in camellia oil and the absorbance of R, SG spectrum. However, there was a relatively weak correlation between the content of palmitic acid, oleic acid and linoleic acid in camellia oil and the absorbance of SD spectrum compared with R, SG spectrum. The accuracy of PLSR model for the content of palmitic acid, oleic acid and linoleic acid in camellia oil built by the full wavelengths was slightly higher than by the significant wavelengths, whose related RC from 0.837 to 0.956 and RP from 0.818 to 0.938, respectively. For the complexity of two models, input variables to the models built by the significant wavelengths were decreased to below 25% compared to that by the full wavelengths. The performance of SG–PLSR model by the full wavelengths was best to test the content of palmitic acid, oleic acid and linoleic acid in camellia oil, whose related RP and RMSEP were 0.938, 0.930, 0.925 and 0.560, 0.438, 0.287, respectively.