Deep learning approach for GPR image classification

Ngoc Quy Hoang, Seonghun Kang, Sang Yeob Kim, Junghee Park, Jong Sub Lee, Hyung Koo Yoon

Research output: Chapter in Book/Report/Conference proceedingChapter


The objectives of this study are to classify ground penetrating radar (GPR) images using a deep learning network trained on raw GPR images and to verify the performance of GPR images treated with white noise. The image dataset includes manually labeled images with three classes: background noise, hyperbola, and manholes. In addition, white noise is introduced into the raw GPR images with different coefficients of variation and new images are used to train the deep learning model. The experimental results reveal that the deep learning model performs well on the raw GPR images, with high classification accuracy. Furthermore, the white noise added to the raw images increases the robustness of the deep learning model. The results of the present study demonstrate the application of a deep learning approach and white noise for classifying GPR images.

Original languageEnglish
Title of host publicationSmart Geotechnics for Smart Societies
PublisherCRC Press
Number of pages4
ISBN (Electronic)9781000992533
ISBN (Print)9781003299127
Publication statusPublished - 2023 Jan 1

Bibliographical note

Publisher Copyright:
© 2023 selection and editorial matter, Askar Zhussupbekov, Assel Sarsembayeva & Victor N. Kaliakin; individual chapters, the contributors.

ASJC Scopus subject areas

  • General Engineering


Dive into the research topics of 'Deep learning approach for GPR image classification'. Together they form a unique fingerprint.

Cite this