To improve the recognition accuracy and real-time performance of transmission line icing monitoring, this paper proposes a classification and recognition method based on intra-class and inter-class distances and adaptive boosting random configuration network (AdaBoost-SCN). First, the phase signals from phase-sensitive optical time-domain reflectometry are jointly evaluated using intra-class and inter-class distance metrics. A scoring strategy is then applied to reduce the dimensionality of the full feature vector, thereby extracting key sensitive features. Subsequently, the AdaBoost-SCN algorithm is employed for icing severity classification and recognition, where the random configuration network serves as the base classifier, and a strong classification model is constructed through iterative optimization. The experimental results demonstrate that the proposed method achieves an average recognition accuracy of 94.7% on the test set, outperforming traditional methods by 2%~5%, while reducing computation time from 0.54 s to 0.32 s.