Aiming at the problem that the existing wine detection technology cannot quickly and efficiently identify the quality of wine, this paper proposes a method to capture the spectral information of wine samples by laser-induced fluorescence technology, which is based on the clustering algorithm without label build and complex tuning. Four wine samples from three brands and two vintages were selected. After being mixed with water at 1∶10, under the premise of the same alcohol content, 100 spectral informations were collected for each wine sample. We used K-means, self-organizing competition network and self-organizing feature mapping neural network (SO-FMNN) to identify wine samples. The experimental results show that the three clustering algorithms have superior performances in spectral information analysis, and the recognition accuracy rate can reach more than 99%. The classification accuracy rate of SO-FMNN is even 100%, the average time is 5.875 s, and it has high robustness and generalization ability. It is verified that the clustering algorithm for wine quality detection is feasible.