Target detection and recognition of ground penetrating radar using morphological image analysis and graph laplacian regularisation

Dong, Jun and Stankovic, Vladimir and Davidson, Nigel; (2021) Target detection and recognition of ground penetrating radar using morphological image analysis and graph laplacian regularisation. In: 2021 Sensor Signal Processing for Defence Conference (SSPD). IEEE, GBR. ISBN 978-1-6654-3314-3 (https://doi.org/10.1109/SSPD51364.2021.9541516)

[thumbnail of Dong-etal-ICSSPD-2021-Target-detection-and-recognition-of-ground-penetrating-radar-using]
Preview
 Text. Filename: Dong_etal_ICSSPD_2021_Target_detection_and_recognition_of_ground_penetrating_radar_using.pdf
Accepted Author Manuscript
Download (265kB)| Preview

Abstract

Ground Penetrating Radar (GPR) is often used for detecting non-intrusively buried targets, in road engineering, manufacturing, and in military fields. Based on transmitting high frequency electromagnetic waves, GPR generates high resolution 3D data of the underground structure enabling accurate and fast target detection. However, after inverse Fourier Transform, the 3D GPR images are often out-of-focus and contain high measurement noise. This calls for advanced signal and image processing methods to improve signal-to-noise ratio, isolate the most discriminative features, and perform target detection and localisation. Using a vehicle-mounted GPR array data provided in the 2020 UDRC GPR data challenge, we show that morphological image analysis and semi-supervised learning via graph Laplacian regularisation can detect different types of targets buried at various depths with very low false alarm rate.

ORCID iDs

Dong, Jun, Stankovic, Vladimir ORCID logoORCID: https://orcid.org/0000-0002-1075-2420 and Davidson, Nigel;
CORE (COnnecting REpositories)