Analyzing Seismic Signal Using Support Vector Machine for Vehicle Motion Detection

A system to process seismic signals of vehicles passing between two sensor stations had been developed and experimented. To evaluate the feasibility of the system before field test with a real vehicle and to support the classification model with artificial data later, the input seismic data were simulated from Green’s method function that accounts only for Rayleigh surface wave. The system using the Machine Learning Classification method SVM to classify data collected from two stations at any time have the state of passed or not. By processing the signal, the system could detect whether the vehicle had passed the crossing line or not with the accuracy of 99.10% for simulated data and 94.22% for experiment data. The experiment and results suggested that processing seismic signals to monitor control lines is feasible.