After pretreatment of 80 samples of maize, interval partial least square (iPLS), combination of interval partial least squares (SiPLS) and successive projections algorithm (SPA) is respectively used to optimize the best wavelength of moisture components, and the correction model is established. The results show that iPLS, SiPLS and SPA method reduces the modeling variables from 700 to 70, 140 and 2, respectively, which occupies 10%, 20% and 0.29% of the whole spectrum. And, the modeling accuracy is even better than that of the 700 full spectral variables. The modeling accuracy of SiPLS and SPA is matched. But the SPA method reduces variables from 700 to 2. The complexity is minimized, and the precision of the model is kept, which show that the SPA method is an effective feature extraction method of wavelength. This research method can be extended to the application of fat, protein and starch components detection of corn.