Apple mealiness is a symptom of internal fruit disorder. Mealiness degrades the quality of apples and reduces their commercial value. Hyperspectral scattering, as a promising technique, combines the advantages of spectroscopy technology and image technology, and can make noninvasive measurement of apple mealiness. A supervised isometric feature mapping (S-Isomap) coupled with support vector machine (SVM) is proposed to detect the mealiness in the apple. S-Isomap is a nonlinear lowering dimension method classifying the dimension reduction of hyperspectral data by SVM. For the unknowned category of the test samples, BP neural network model combined with SVM is used to get the corresponding testing precision. The classification results from S-Isomap-SVM are compared with those obtained using the traditional SVM and Isomap-SVM. The results show that the accuracy of the calibration models obtained with the S-Isomap is higher than that of others.