Abstract: The spectral data of dyed navel oranges was achieved by near infrared spectrum instrument (NIRS). The data was preprocessed by different means and analyzed with principal component analysis (PCA). A PLS model was established based on this,and the performance of the model was evauated according to root mean squared error of calibration (RMSEC) and correction coefficient (R²). Results show that:It appeared to provide PCA could be used to identify the dyed navel oranges,the recognition rate achieved 94%. RMSEC of the PLS regression model established based on PCA was 0.26,R² was 0.96,the model was well. It was feasible to identify the dyed orange by near infrared spectroscopy,which would provide a new and nondestructive method for the identification of dyed orange.