In order to solve the problem of identification moldy core of apples from the surface, a quick and non-destructive detection method was proposed based on visible/near infrared transmission energy spectroscopy. Visible/near infrared transmission energy spectra of 200 apples were collected in the wavelength range of 200 ~ 1100nm. Totally 140 samples were used for the calibration set, and 60 samples for the validation set. Smoothing method and multiple scattering correction were used to preprocess the original spectra. Totally 12 characteristic wavelengths and 9 principal components were selected by successive projections algorithm (SPA) and principal component analysis (PCA), respectively. Partial least squares discriminant analysis, error back propagation artificial neural networks, and support vector machine (SVM) measurement model were established based on SPA and PCA, respectively. The results showed that the best model was PCA—SVM, and its recognition accuracy rate reached 99.3% for the calibration set and 96.7% for the validation set. The models established based on SPA and PCA were much simpler than those based on full spectra, since the numbers of input variable of them were only about 0.99% and 0.74% of that of full spectra, respectively. The results showed that the method was available and had high identification accuracy. Meanwhile, the results would provide theoretical basis for the research and development of on-line detection of internal quality in apples and portable moldy core apple detector.