Acta Photonica Sinica, Volume. 54, Issue 1, 0106002(2025)
Distributed Fiber Optic Acoustic Sensing System Based on Fading Mask Autoencoder and Application in Water Navigation Security Events Identification
Figures & Tables(15)
Fig. 8. Data visualization of the four types of ship events in waters
Fig. 9. Mean comparison of ten cross-validation training results for two models：FMAE and MAE
Fig. 10. Comparison of training results for the four models，FMAE/ViT/SENet/ResNet
Table 1. Data sets used for the four types of shipping events in waters
View tableView in ArticleTable 1. Data sets used for the four types of shipping events in waters
Event type （I）drone ship （II）boating （III）speedboat （IV）no event Total Train set 16 600 18 000 17 000 18 000 69 600 Test set 2 000 2 000 2 000 2 000 8 000 Total 18 600 20 000 19 000 20 000 77 600
Table 2. Comparison of performance measures between FMAE and MAE models
View tableView in ArticleTable 2. Comparison of performance measures between FMAE and MAE models
Average training
accuracy/%
Average training
loss
Average test
accuracy/%
Average precision Average F1-score FMAE 98.34 0.109 4 93.01 0.934 5 0.929 6 MAE 93.44 0.204 4 86.56 0.893 1 0.858 7
Table 3. Performance indicators of the four models，FMAE / ViT / SENet / ResNet
View tableView in ArticleTable 3. Performance indicators of the four models，FMAE / ViT / SENet / ResNet
I II III IV Recall Accuracy FMAE 0.786 0 0.944 5 0.990 0 1.000 0 0.930 1 ViT 0.379 5 0.994 5 1.000 0 1.000 0 0.843 5 SENet 0.576 5 0.946 5 1.000 0 1.000 0 0.880 7 ResNet 0.707 0 0.772 5 1.000 0 1.000 0 0.869 8 Precision Average precision FMAE 0.934 1 0.818 8 0.994 9 0.990 1 0.934 5 ViT 0.985 7 0.626 0 0.974 1 1.000 0 0.896 4 SENet 0.915 0 0.711 9 0.961 0 1.000 0 0.897 0 ResNet 0.769 3 0.727 4 0.997 0 0.984 2 0.869 4 F1-score Average F1-score FMAE 0.853 6 0.877 2 0.992 5 0.995 0 0.929 6 ViT 0.548 0 0.768 3 0.986 9 1.000 0 0.825 8 SENet 0.707 3 0.812 6 0.980 1 1.000 0 0.875 0 ResNet 0.736 8 0.749 2 0.998 5 0.992 0 0.869 1
Table 4. Inference time of four models，FMAE / ViT / SENet / ResNet
View tableView in ArticleTable 4. Inference time of four models，FMAE / ViT / SENet / ResNet
FMAE ViT SENet ResNet Inference time 10.75 ms 6.98 ms 16.96 ms 17.95 ms
Tools
Get Citation
Copy Citation Text
Miao YU, Yutong HE, Tianying CHANG, Hongliang CUI, Suihhu DANG, Liangping XIA, Liming LIU, Zichuan YI, Xinjian PAN, Qingguo GAO. Distributed Fiber Optic Acoustic Sensing System Based on Fading Mask Autoencoder and Application in Water Navigation Security Events Identification[J]. Acta Photonica Sinica, 2025, 54(1): 0106002
Paper Information
Category: Fiber Optics and Optical Communications
Received: Jun. 12, 2024
Accepted: Aug. 19, 2024
Published Online: Mar. 5, 2025
The Author Email: CUI Hongliang (hl.cui@siat.ac.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20