A near infrared spectrum preprocessing method based on Daubechies wavelet threshold is proposed. It is used to process the near infrared spectra of two kinds of apples. In the method, the normalized Correlation Coefficient (NC) and Peak Signal to Noise Ratio (PSNR) are selected as a quantitative analysis index. Compared with the Vector Normalization (VN) and Standard Normal Variate (SNV), the method has obvious advantages. It not only can effectively remove noises, but also can retain the details of spectral characteristics. So, the robustness of modeling and the accuracy of the established model can be improved in the subsequent spectral analysis.