Abstract: To understand corn straw carbohydrate composition, a corn straw cellulose and hemicellulose contents model was developed using Near-infrared Spectroscopy (NIRS). An ideal model was obtained after multiple trial-and-error prediction/validation processes, using spectrum analysis together with partial least squares (PLS) method. Based on this model, influences of different pretreatment methods on corn straw cellulose and hemicellulose contents were investigated. A correlation coefficient of 0.909 was achieved. Compared with the experimental data, the values predicted by this NIRS model have mean relative errors of 2.34% and 2.13% for cellulose and hemicellulose contents, respectively. This model is capable of accurately and quickly analyzing corn straw cellulose and hemicellulose contents at industrial scale, which is beneficial to the utilization of biomass resources and the biomass conversion process.