Trains Detection Using State of Polarization Changes Measurement and Convolutional Neural Networks

Abstract

Fiber optic infrastructure security is of growing interest. The current distributed sensor systems are robust and expensive solutions, and their practical applications are uncommon. Research into simple and cost-effective solutions based on changes in the state of polarization is crucial. This paper expands the use of a vibration sensor based on the sensing of rapid changes in the state of polarization (SOP) of light in a standard single-mode optical fiber by using a convolutional neural network to detect trains running along the optical fiber infrastructure. It is a simple system that determines ongoing events near the optical fiber route by simply determining the signal boundaries that define the idle state. By using a neural network, it is possible to eliminate the distortion caused by the temperature changes and, for example, to improve detection in the the zones where the vibrations are not strong enough for a simple threshold resolution.