Abstract: Abstract: Rice is rich in starches, proteins and carbohydrates, and when it is polluted by fungus, it is easy to become decayed and hence produces some poisonous substances for human bodies. Once moldy rice goes into the circulation market, human health will suffer from serious hazard. Therefore, how to effectively detect fungus in rice has become a fundamental work of guarantying food security. At present, the detection of moldy rice mainly depends on artificial qualitative analysis, which means that detectors discriminate fungus in rice according to some physical indices such as color and aroma. The detection precision of the mentioned methods mostly depends on the knowledge or experience of operators and the indication of statistic tools chosen by operators, which will bring out artificial errors inevitably. The fatty acid content is an important indicator of fungus information in rice. In order to solve these problems presented in the traditional way such as destruction, time consuming and low efficiency, a non-destructive detecting method for fatty acid content in rice using high-spectral technologies was proposed in this paper. In the research, rice samples for 4 different storage periods by means of artificial cultivating were selected as study objects, and spectral information and fatty acid content were detected through high-spectral measurement and physical and chemical experiment. The spectral data obtained were preprocessed using the Savitzky-Golay (SG) smoothing and the first derivation (FD) method, and the characteristic spectrum that indicated the variations of fatty acid content was selected by the successive projections algorithm (SPA). The prediction model of fatty acid content in rice based on spectral reflectance was built by the regression analysis method, and the prediction effect was evaluated by comparing different preprocessed methods. Experimental results indicated that 14 and 10 spectral characteristic wavelengths, which were from the original spectral data after the SG smoothing and the FD preprocessing, were optimized and selected according to the SPA. The quality of modeling and prediction effect for fatty acid content in rice showed that the SG-SPA-MLR (multivariable linear regression) method was superior to the FD-SPA-MLR method. The correlation coefficient of cross-validation (Rcv) and the root mean square error of cross-validation (RMSECV) for the SG-SPA-MLR model were 0.9419 and 11.9646 mg/100 g respectively at the model correction stage, while the correlation coefficient of prediction (Rp) and the root mean square error of prediction (RMSEP) were 0.9366 and 12.3550 mg/100 g respectively at the stage of the model prediction. The optimal model showed a good prediction ability in fatty acid content of rice during different storage periods. In summary, the results have indicated that it is feasible to non-destructively predict fatty acid content variation in rice applying high-spectral technologies, and can be used as the reference for the rapid detection of fungus stress in rice in the future.