On the basis of extracting all the characteristic peaks of laser-induced breakdown spectroscopy (LIBS), an effective tea classification model is established based on support vector machine. The effective LIBS spectral data (190-720 nm) of fifteen tea samples are collected, and the spectra are preprocessed by window translation smoothing and peak shift function correction. Combined with principal component analysis for dimensionality reduction, the recognition rate of green tea, black tea and white tea is 98.3%. Different varieties of tea in the same species also achieve good recognition. The research results show that LIBS has a good prospect in the rapid identification of tea varieties.