Classifying GPR images using convolutional neural networks

This thesis focused on classifying GPR cylinders' B-scans according to their depth, size, material, and the dielectric constant of the underlying medium using four different architectures of convolutional neural networks. Two CNNs were newly proposed for this study, while the other two were used by other authors. These CNNs were trained using a couple of adjusted training options including initial learning rate, learn rate drop factor, and learn rate drop period; which had a positive impact on a part of the used models, while the option maximum number of epochs worked good with all of the used models. Results show that the first newly proposed CNN showed a superior performance due to the use of a deep network with a large amount of small filters. Using this model, it was found that the best results were carried out when GPR B-scans were classified according to the cylinders' materials.

Grantor dc:publisher

University of Tennessee at Chattanooga

Author dc:creator

• Almaimani, Maha

Contributors dc:contributor

• Wu, Dalei
• Liang, Yu; Yang, Li
• College of Engineering and Computer Science

• Ground penetrating radar
• Neural networks (Computer science)

Licence

• https://rightsstatements.org/page/InC/1.0/?language=en

Language dc:language

English, eng

Repository record dc:identifier

https://scholar.utc.edu/theses/544

OAI identifier oai:identifier

oai:scholar.utc.edu:theses-1697