The invention provides a hail recognition method under multi-dimensional radar data based on semi-supervised learning, and the method comprises the steps: S1, obtaining a labeled sample set, randomly extracting a supervised sample set, a rainstorm sample training set and a hail sample training set, obtaining an unlabeled data set, and randomly dividing the unlabeled data set into q parts of first samples; s2, calculating a clustering center of each cluster training set; s3, clustering and dividing one first sample into corresponding clusters, and updating a clustering center; s4, performing iteration to obtain the clustering center of each cluster and the confidence coefficient of the corresponding cluster at the moment; s5, repeating the steps S2-S4 on the supervision sample set to obtain supervision confidence of the supervision sample set for each clustering center, and classifying the data into corresponding clusters; s6, judging whether the first sample is updated into the cluster or not, and repeating the steps S2-S6 until the first sample is processed; and S7, inputting the optimal clustering center as a recognition model, obtaining the confidence coefficient of each sample to each cluster, and performing classification. According to the method, the hail identification accuracy is effectively improved, and the false alarm rate is reduced.