A characteristic wavelength for measuring flesh browning of apple was put forward based on FT-NIR spectroscopy. Cluster analysis algorithm based on Euclidean distance was applied to selection of representative samples. The multivariable analysis concluding PLS and MLR were applied to build the regression models and seek the characteristic wavelengths of apple flesh browning. The loading weights from PLS model found the sensitive flesh browning wavelengths. The significance of these wavelengths was by ANOVA, and seven wavelengths (1132, 1411, 1429, 1494, 1580, 1879, and 2151nm) were selected with statistical significance at 95% confidence level. These wavelengths were strongly related with apple flesh browning by MLR models evaluated with high prediction correlation (r=0.937). These suggest the model of apple flesh browning was reliable with good predict ability and can meet the requirement to quick determination of flesh browning of apples.