The perimeter security alarm system, with the core of vibration sensor based on FBG optical fiber invented by Wuhan University of Technology optical fiber research center, uses the changes of optical fiber grating stress wavelength to detect intrusion. As FBG wavelengths change when under stress, the intrusion detection is identified by the combination of signals time domain and frequency band domain. However, it is difficult to classify the data of signals time domain and frequency band domain because of the complexity of the actual conditions and the diversity of interference and intrusion. Therefore, applying SVM to train and process the classification features extracted from the various data acquisition can help to obtain the corresponding support vector and weighting coefficient. As a result, the effective classification of the field test data can be achieved by using SVM template method.