The classified feasibility of pear jujube with normal, rot and disease defect fruit by using hyperspectral imaging technology was analyzed. A best mathematical model was established by treating first derivative as the best pre-processing which was compared with other different kinds of proceeding methods. The stepwise discriminant analysis and least square support vector machines (LS—SVM) were applied to build the full-wave band, approximation coefficients and principal components models, respectively. The results indicated that the stepwise discriminant analysis model was more suitable for classifying the three different kinds of pear jujube samples. The average correct recognition rate was 99.12%.