In order to improve the identification accuracy of moldy degrees of maize samples using hyperspectral, the identification effects of moldy maize at the full wavelength and the characteristic wavelength were investigated in this study, respectively. Firstly, the hyperspectral data of 250 moldy maize samples were obtained by hyperspectral image acquisition system, and the standard normal variate (SNV) and multiplicative scatter correction (MSC) were employed to preprocess the original data; and then the MSC was adopted by comparing the results of the preprocessed and non-preprocessed data. Secondly, nine characteristic wavelengths were selected by using the partial least squares regression coefficients. Finally, Fisher discriminant analysis (FDA) was used to analyze the training set at full wavelengths and characteristic wavelengths, and examined by the corresponding test set. The FDA results showed that the accuracy of the training set and test set were 97.71% and 97.33% for full wavelengths case, respectively, and were 100.00% and 98.67% for characteristic wavelengths case, respectively. The accuracy of the training set and test set at 9 characteristic wavelengths were 100.00% and 98.67% respectively. The research findings showed that the characteristic wavelengths could be used to represent the moldy degrees of maize samples, which was helpful to improve the identification correct rate of moldy maize. Moreover, the research findings might provide a reference for identifying the other agricultural products using hyperspectral technology.