The traditional support vector machine has two faults: low classification accuracy and poor timeliness. In order to obtain support vector machine (SVM) with high accuracy and efficiency, the parameter optimization of SVM with mixed kernels based on mutative scale chaos particle swarm optimization (MSCPSO) was presented. This model was used to predict the growth stage of lettuce leave, which was consist of seedling stage, tillering stage and mature stage, and N content levels of three growth periods respectively. The prediction accuracy achieved to 91.51%, 85.38%, 82.59% and 81.26%. Compared with the traditional particle swarm optimization mixed nuclear SVM classifier and mutative scale chaos particle swarm optimization RBF_SVM classifier, the proposed classifier model showed higher classification accuracy and timeliness.