skip to main content
Idioma:
Primo Search
Clear Search Box
Search in: Busca Geral
Busca Avançada
Busca por Índices

Learning Spatial-Temporal Features of Fiber-Optical Data with Multi-Scale Double Dynamic Network for Human Intrusion Detection

Zhao, Shuo ; Guo, Zhongwen ; Cheng, Xu ; Jiang, Sining ; Zhao, Wenchang ; Wang, Hao

IEEE transactions on instrumentation and measurement, 2023-01, Vol.72, p.1-1 [Periódico revisado por pares]

New York: IEEE

Texto completo disponível

Citações Citado por
  • Título:
    Learning Spatial-Temporal Features of Fiber-Optical Data with Multi-Scale Double Dynamic Network for Human Intrusion Detection
  • Autor: Zhao, Shuo ; Guo, Zhongwen ; Cheng, Xu ; Jiang, Sining ; Zhao, Wenchang ; Wang, Hao
  • Assuntos: Artificial neural networks ; Computer architecture ; deep learning ; Fiber optics ; Intrusion detection ; Intrusion detection systems ; layer skipping ; Machine learning ; multi-scale ; Optical communication ; optical fiber ; Optical fibers ; OTDR ; Spatial data
  • É parte de: IEEE transactions on instrumentation and measurement, 2023-01, Vol.72, p.1-1
  • Descrição: Intrusion event detection plays a crucial role in safeguarding national territory and large-scale critical facilities. However, modern detection methods often fail to incorporate spatial information and adequately extract multi-scale information, leading to a decrease in detection accuracy. In this paper, a human intrusion detection method based on distributed optical fiber sensing systems is proposed. Firstly, a linear regression-based data preprocessing approach is proposed to identify important features in fiber-optical data and rearrange the data in a one-dimensional format according to spatial-temporal relationships. Then, a multi-scale deep neural network architecture with double dynamic layer skipping mechanism that enables the network to automatically retain crucial layers and concatenate multi-scale features is proposed for capturing both small-scale details and large-scale long-term regularities in signal changes, leading to efficient extraction and utilization of the temporal and spatial signal information of fiber-optical data. A fiber-optical dataset of human intrusion events is also collected to evaluate the effectiveness of this method, achieving intrusion detection with an accuracy of 99.49%.
  • Editor: New York: IEEE
  • Idioma: Inglês
Voltar para lista de resultados

  • Realização: ABCD-USP USP   Logos de Redes Sociais: Freepik

Buscando em bases de dados remotas. Favor aguardar.