Abstract: This study established a Support Vector Machine (SVM)-based radar classification model of hydrometeor under the T-matrix based radar detection model of hydrometeors. Through normalizing data in the first place, it is also considered that data among polarization parameters are non-linear. Therefore, the study chose radial basis function as the kernel function of non-linear SVM and used Particle Swarm Optimization (PSO) to obtain the optimal kernel function parameters C and γ, so as to achieve higher accuracy of hydrometeor classification. The prediction accuracy of the established SVM-based radar classification model of hydrometeor reached more than 80% at X band and close to 95% at S band at all elevations. Further analysis shows that the prediction accuracy of hydrometeor classification can reach 97.3%, while the misjudgement is only 2.7%, when the prediction types of hydrometeor are the same with multi-wavelength joint observations. In conclusion, the establisehd SVM-based radar classification model of hydrometeor could improve both the ability to classify the hydrometeors in convective weather and the ability for early warning and forecasting of disastrous weather by dual linear polarization radars.