社団法人　物理探査学会
第143回(2020年度秋季)学術講演会

講演要旨(和文) 講演要旨(英文) Vehicle classification used to depend on photos or videos sounds. However, these techniques usually facing criticism due to vaulting the driver's privacy and exposing his identity. Using seismic signals for vehicle identification was a difficult task as a result of various noises even with the frequency information. We consider the similarities between speech recognition and vehicle classification based on the seismic signal we used different Artificial Intelligence (AI) techniques to learn the special features of 3 different size vehicles (Bus, Car, Motorcycle). This study investigates the application of Deep Neural Networks (DNN), Convolutional Neural Networks (CNN), and Recent Neural Networks (RNN) to classify vehicles class by using a single component seismic record that. The Neural Networks were trained on more than 3000 unprocessed seismic records and achieve the best accuracy (99.1%) with CNN. Our algorithm can be used for traffic monitoring and security purposes without vaulting driver's privacy.