Fig 1.
Overall experimental workflow diagram.
Fig 2.
Working principle diagram of φ-OTDR.
Fig 3.
Schematic diagram of the scattered light principle.
Fig 4.
2-D CNN network architecture diagram.
Fig 5.
Network architecture diagram of the 1-D CNN designed in this paper.
Fig 6.
Hyperparameter Sensitivity (Architecture × Optimization, Macro-F1).
Table 1.
Three types of search result statistics table.
Fig 7.
Network training flowchart.
Table 2.
Refer to the statistical comparison and ablation study results table.
Fig 8.
Visualization results of separability before and after t-SNE fusion.
Table 3.
Comparison table of Fusion strategies.
Table 4.
Dataset of Six Simple Human Activities.
Fig 9.
Field data collection schematic.
Fig 10.
PSD overlay and frequency band energy bar comparison chart.
Fig 11.
Scatter plots of kurtosis, zero-crossing rate, and AC decay.
Fig 12.
Mean graph of spectral kurtosis curve.
Fig 13.
Pipeline intrusion event data collection.
Fig 14.
Original data label distribution chart.
Table 5.
Specific parameter table of the network.
Fig 15.
Distribution of each feature.
Fig 16.
Data dimension reshaping diagram.
Fig 17.
Time series chart of different samples.
Fig 18.
Visualization of samples with label 0.
Fig 19.
Model prediction chart before incorporating transfer learning.
Fig 20.
Test results ROC curve diagram.
Fig 21.
Test Results Heat Map.
Fig 22.
Oil and gas pipeline intrusion data test diagram.
Fig 23.
Comparison chart of recall, F1 score, precision, and support for four comparative experimental algorithms.
The four comparative experimental results are integrated into Table 6 as shown below.
Table 6.
Four comparative experimental result charts.
Fig 24.
Comparison chart with various deep learning algorithms.
Table 7.
Compared with a variety of deep learning algorithms.