In order to distinguish the color discrepancy of the same color class, the method of detection based on spectrophotometer and support vector machine is proposed. First, spectrophotometer is utilized to obtain the CIE chromaticity coordinates of each point in the glasses and other optical data for establishing data sample set. Then selecting the feature parameters and establishing the training set are used for creating SVM model. Followed, the different types of kernel functions and parameters are compared to choose the best, so that they make the model optimizing. Finally, the model is used to work on the prediction set, which detecting and distinguishing the dissimilarity. The experimental results show that this method can be efficiently and accurately detected the discrepancy of color of the glass.