Distributed Optical Fiber Intrusion Detection by Image Encoding and SwinT in Multi-Interference Environment of Long-Distance Pipeline

External intrusion incidents pose a severe threat to pipeline security in energy transportation. In response to this, distributed optical fiber sensing technology has been widely studied in the field of safety monitoring in recent years. However, the diversity of the environment along the long-distance pipeline makes the vibration signal complex and changeable, which significantly limits the recognition accuracy in practical applications, resulting in numerous false positives. To address the above issues, we transform intrusion detection into a multiclass classification problem to identify intrusion events. In this study, a scheme of image encoding combined with shifted windows transformer (SwinT) model in computer vision is proposed for pattern recognition. Specifically, the timing signals collected by the distributed vibration sensing (DVS) system are transformed into 2-D images. The correlation and time dependence between sampling points are strengthened in image encoding, and the window and shifted window design of SwinT are used for multiscale feature extraction. Moreover, the focus loss function is introduced to attenuate the impact of the class imbalance issue in the actual scene. Extensive experiments verify the superiority of our proposed method in terms of various evaluation indicators, demonstrating that the model can be deployed online for energy pipeline safety.