Signal Classification of Distributed Fiber Optic Sensor System

The application of machine learning methods has attracted widespread interest for several years. One possible area where the need for these techniques is high is in security systems. In this final work, the task of classifying the signals of the monitoring system received from a phase-sensitive optical reflectometer, the cable of which is laid along the perimeter of the protected object, is posed. Several machine learning models have been studied and proposed to solve the problem. Many of the methods that have been studied are applicable to computer vision problems and are excellent for this issue. In the course of this work, data preprocessing was performed for their subsequent use in models. The result was a comparative analysis of the quality and accuracy of various algorithms from machine learning. Signals issued by the original system as alarms were taken as input data. Convolutional neural network architecture has also been proposed as a solution to the classification problem. The highest achieved classification accuracy among the implemented algorithms was observed in the neural network and exceeds 98%