| In order to determine the contents of lignin, cellulose and hemicellulose more rapidly and accurately, the near-infrared reflectance spectroscopy(NIRS) models of fiber composition in corn stalk were constructed with 200 maize inbred lines as experimental materials. The results showed that the NIRS models of lignin, cellulose and hemicellulose content in maize stalk established in the spectral regions of 4017.94~8053.28, 4017.94~8067.89 and 4027.08~8928.20 were the best. The calibration model was established by partial least square regression (PLS), and the corrected correlation coefficients (RMRSQ) of lignin, cellulose and hemicellulose were 0.9329, 0.9251 and 0.9265, and the corrected standard deviation (RMSEC) were 1.57, 1.68 and 1.18, respectively. In order to verify the model, 30 corn stalk samples were selected as test set. The external correlation coefficients (RSQ) of lignin, cellulose and hemicellulose were 0.9389, 0.8911 and 0.9050, respectively, and the prediction standard deviations (SEP) were 1.57, 2.14 and 1.49, respectively. The models were also cross-validated with 30 stalk samples, and the cross-validation correlation coefficients (1-VR) were 0.8973, 0.9442 and 0.8918, respectively, and the standard deviation of cross validation (SECV) ware 1.87, 2.32 and 1.43, respectively. The results showed that the models have good quality and can determine the contents of lignin, cellulose and hemicellulose in corn stalk quickly and accurately, which will provide technical support for corn lodging resistance breeding research.