Enhanced Underground Object Detection with Conditional Adversarial Networks  
Author Will Rice


Co-Author(s) Maxwell Omwenga; Dalei Wu; Yu Liang


Abstract To augment training data for machine learning models in Ground Penetrating Radar (GPR) data analysis, this paper focuses on the generation of realistic GPR data using Generative Adversarial Networks (GAN). An innovative GAN architecture is proposed for generating GPR B-scans, which is, to the author’s knowledge, the first successful application of GAN to GPR data. As one of the major contributions, a novel loss function is formulated by merging frequency domain features with time domain features. To test the efficacy of generated B- scans, a real-time object classifier is proposed to measure the performance gain derived from augmented B-Scan images. The numerical experiment illustrated that, based on the augmented training data, the proposed GAN architecture demonstrated a significant increase (from 82% to 98%) in the accuracy of the object classifier.


Keywords Ground Penetrating Radar, Generative Adversarial Networks, Underground Object Detection, Classification
    Article #:  DSIS19-59
Proceedings of ISSAT International Conference on Data Science & Intelligent Systems
August 1-3, 2019 - Las Vegas, NV, U.S.A.