The effective identification of fiber optical vibration signals is an important basis for ensuring the operation of the fiber-optical early warning system for oil and gas pipelines. To mitigate the lack of a single classification method in traditional fiber optical vibration signal detection, this paper proposes a fiber optical vibration signal recognition and classification algorithm using AdaBoost ensemble learning. First, we analyzed and studied the characteristics of five fiber optical vibration signals and selected sample entropy, energy distribution, and bandwidth as the three-dimensional feature vectors. Next, this information was sent to the decision tree, support vector machine (SVM), and AdaBoost classification algorithm with the decision tree as the base classifier for training and recognition. Second, the obtained models were optimized and evaluated by cross-validation. Finally, the obtained models were compared. The experimental results show that the AdaBoost ensemble learning algorithm with a decision tree as the base classifier effectively identifies different vibrations and has certain significance for identifying vibration signals from different sources in the fiber-optical warning.