DVS-Based Vibration Spectrum Analysis for Optical-Fiber Path Matching

Rational planning of fiber-optic paths can significantly mitigate network damage caused by construction activities and natural disasters while enhancing fault tolerance capabilities. However, current methods for fiber-optic path matching primarily depend on expert experience and extensive data support. As the complexity of fiber-optic network construction grows, these methods become increasingly constrained due to the absence of a unified framework. To tackle this challenge, this study employs Distributed Optical Fiber Vibration Sensing (DVS) technology to precisely capture fluctuation data in communication scenarios and introduces a multi-scale feature-enhanced nearest-neighbor matching algorithm (ERPM). The ERPM algorithm leverages a multi-scale convolutional neural network to extract hierarchical feature information and employs a cross-attention mechanism to refine the feature vectors. It then utilizes a nearest-neighbor matching algorithm to compute vector similarity, effectively addressing displacement and deformation challenges in path matching. To validate the algorithm’s efficacy, a 7-day experimental measurement was performed on two fiber-optic cables spanning 8 kilometers. The results indicate that the ERPM algorithm achieved a 98% accuracy rate in path matching, representing an 8% improvement over the existing DIS method that integrates neural networks. This demonstrates the algorithm’s superiority and potential value in practical applications.