EUVIMEDEUVIMED entdecken
Crossref

Spatial--Temporal Feature Fusion-Assisted Brillouin Frequency Shift Extraction in BOTDA With Enhanced Accuracy and Efficiency

Caojun Zhang, Yichun Li, Yu Chen, Weiyu Pan, Zhifeng Wang, Heming Wei, Mengshi Zhu, Fufei Pang, Liang Zhang

IEEE Sensors Journal · 15.07.2026 · 26 (14) · 20857-20864

DOI: 10.1109/jsen.2026.3696750 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Sensors Journal, erschienen am 15.07.2026. Verfasst von Caojun Zhang und 8 weiteren Autor:innen. Die Arbeit nennt 25 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Shanghai Pujiang Program, China Postdoctoral Science Foundation.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Accurate extraction of the Brillouin frequency shift (BFS) is crucial for distributed fiber sensing based on Brillouin optical time-domain analysis (BOTDA). However, practical measurements are severely degraded by inherent noise, limiting precision and reliability. This article proposes a novel deep-learning framework for robust BFS extraction, uniquely combining spatial–temporal joint denoising with attention mechanisms. Our approach processes the entire noisy Brillouin gain spectrum (BGS) trace simultaneously across both spatial and temporal dimensions, capturing comprehensive noise characteristics and signal dependencies. Crucially, a channel attention mechanism (CAM) is embedded within the denoising network architecture, enabling it to dynamically focus computational resources on the most informative spectral regions and spatial segments critical for accurate BFS estimation while suppressing irrelevant or noisy components. Subsequently, an integrated framework incorporating convolutional neural networks (CNNs), long short-term memory (LSTM), and multi-head attention (MHA) is proposed. This framework seamlessly fuses capabilities for convolutional feature extraction, temporal dependency modeling, and frequency-domain attention guided decision making, achieving end-to-end, high-precision BFS extraction. Compared with the traditional Lorenz curve fitting (LCF) method, the proposed method improves the BFS extraction accuracy by 8.36% in simulations and 3.54% in experiments, while the extraction speed is accelerated by a factor of 63 in simulations and 15 in experiments. The proposed method achieves a substantial improvement simultaneously in both the measurement accuracy and data processing speed, paving the way for reliable, and efficient distributed fiber sensing.

Förderung

  • National Natural Science Foundation of China
  • Shanghai Pujiang Program
  • China Postdoctoral Science Foundation
  • Shanghai Professional Technology Platform
  • Higher Education Discipline Innovation Project

Zitieren

APA

Zhang, C., Li, Y., Chen, Y., Pan, W., Wang, Z., Wei, H., Zhu, M., Pang, F., & Zhang, L. (2026). Spatial--Temporal Feature Fusion-Assisted Brillouin Frequency Shift Extraction in BOTDA With Enhanced Accuracy and Efficiency. IEEE Sensors Journal, 26(14), 20857-20864. https://doi.org/10.1109/jsen.2026.3696750

Vancouver

Zhang C, Li Y, Chen Y, Pan W, Wang Z, Wei H, et al. Spatial--Temporal Feature Fusion-Assisted Brillouin Frequency Shift Extraction in BOTDA With Enhanced Accuracy and Efficiency. IEEE Sensors Journal. 2026;26(14):20857-20864. doi:10.1109/jsen.2026.3696750

BibTeX
@article{Zhang2026,
  author = {Zhang, Caojun and Li, Yichun and Chen, Yu and Pan, Weiyu and Wang, Zhifeng and Wei, Heming and Zhu, Mengshi and Pang, Fufei and Zhang, Liang},
  title = {Spatial--Temporal Feature Fusion-Assisted Brillouin Frequency Shift Extraction in BOTDA With Enhanced Accuracy and Efficiency},
  journal = {IEEE Sensors Journal},
  year = {2026},
  volume = {26},
  number = {14},
  pages = {20857--20864},
  doi = {10.1109/jsen.2026.3696750}
}
Kostenlos starten →