Color-difference was presented to assess the lycopene content conveniently and non-destructively. Due to excessive affecting factors and strong correlation among the parameters in the process, the support vector machine (SVM) was used to set up the predict model. The selection and simplification of the feature parameters was discussed. A compound optimal objective function based on Akaike information criterion (AIC) was constructed. The particle swarm optimization (PSO) algorithm was used to search the optimal value of the objective function and enhance the efficiency. The predictable method had good performance in assessing the lycopene content of different maturity stages.