Abstract: The model for rapid nondestructive identification of quality of walnut kernel was established by near infrared hyperspectral imaging technology combined with partial least square discriminant analysis (PLS-DA). The spectral data of walnut kernels with different quality were collected in the full wavelength range of 900-1 700 nm, and the average spectra were used as the original spectra. The original spectral data was pretreated by standard normal variate. The dimension of the original spectral data was reduced by principal component analysis, and 11 characteristic wavelengths were extracted, including 970, 1 151, 1 210, 1 215, 1 256, 1 309, 1 340, 1 379, 1 389, 1 404, 1 460 nm. Two PLS-DA classification models were established based on full spectra and characteristic wavelengths. The results showed that the prediction accuracy of model under full spectra condition on the calibration set and verification set was the highest, reaching 100%, while that of model under characteristic wavelengths condition on the same data sets decreased slightly, reaching 99.3%. The prediction accuracy of the both models on the testing set was 100%.