Exploration of Perimeter Security Intrusion Recognition Method Based on DAS and GASF-MSCNN

To improve the perimeter security intrusion recognition accuracy, in this article, an intrusion identification model based on the Gramian angular summation field (GASF) and the multiscale convolutional neural network (MSCNN) is proposed, which is based on the distributed acoustic sensing (DAS). This model converts the time-domain vibration signals of optical fibers collected by DAS into 2-D GASF images. Then, MSCNN is used for feature extraction and classification of the GASF images, resulting in a high-classification accuracy. By collecting and classifying the vibration signals of optical fibers for the scenarios of no intrusion, striking, shaking, trampling, and shoveling through DAS, the experimental results reveal that the classification accuracies of this model for the above five types of vibration events are 100%, 98.82%, 98.15%, 98.32%, and 97.76%, respectively. Compared with the classical convolutional neural network (CNN), long short-term memory (LSTM) network, and temporal convolutional network (TCN), the average accuracy is improved by 5.6%, 15.28%, and 8.42%, respectively. This work introduces an innovative technical methodology that can effectively improve the perimeter security intrusion recognition accuracy.