The rapid and nondestructive detection method of egg freshness was investigated based on the near-infrared spectroscopy technology combined with the wavelength selection method in this study. We promoted using the egg yolk index as the target to estimate the freshness of eggs. The results showed that the models established for pH value and Hough units were unacceptable and the egg yolk index detection models were the best. By applying the successive projections algorithm (SPA), 12 wavenumbers were chosen out, i.e., 4 188.6，4 593.6，4 855.9, 5 311.0, 5 376.6, 5 935.8, 6 306.1, 7 243.3, 7 328.2, 7 343.6, 8 130.4 and 8 531.5 cm^-1. According to the index of these12 variables , three egg yolk models, including partial least squares regression (PLSR), principal component regression (PCR) and stepwise multivariate linear regression (SMLR), were established respectively. The results of these three models were basically similar, and the SMLR were the best among them. The prediction correlation coefficient (r_pre) and the root mean square error of prediction (RMSEP) were 0.950 and 0.030 of SPA-SMLR model, respectively. Our results indicated that detecting the egg yolk index by using the near-infrared spectroscopy technology was practicable. Moreover, the performance could be improved by using the feature wavelengths as the input. This study will provide a theoretical basis and new clue for the further development of the potential effective and nondestructive detective equipment in egg freshness grading.