The transmittance/reflectance NIR spectra of fresh apple juice were collected by immersed fiber at four optical path of 5mm, 10mm, 15mm, and 20mm. The NIR spectra were used to predict quantitatively the soluble solid contents (SSC) and pH value of apple juice. The results show that the optimal optical path lengths for SSC and pH are different, which are 5mm and 20mm, respectively. In order to consider the concentration ranges of different components and their op- timal path lengths, multi-optical path spectra were used to construct the model to improve the model quality. The method for extracting information from multi-optical-path spectra was studied. The method based on directly unfolded spectra could effectively use the spectral information, but the model became more complex and the calculation time was very long. Two methods based on wavelet transformation were proposed to extract the information. They can extract the multi-optical path spectra information effectively, and at the same time the calculation time decreases significantly and the model is simpli- fied. One of the methods that uses the approximate wavelet coefficient of the unfolded spectra to construct model has the best performance: SECV for SSC and pH value are 0.4761° Brix and 0.0779, respectively.