Deep learning-based 3D imaging study of ground-penetrating radar
Physica Scripta · 06.03.2026 · 101 (9) · 095003
DOI: 10.1088/1402-4896/ae4790 ↗Kurz gefasst
Fachartikel in der Zeitschrift Physica Scripta, online erschienen am 02.03.2026. Verfasst von Yongchao Yin und 4 weiteren Autor:innen. Die Arbeit nennt 26 Literaturangaben. Gefördert von: Science and Technology Support Project for the Construction of Ordos National Innovation Demonstration Zone for the Sustainable Development Agenda, Youth Fund of the Inner Mongolia Natural Science Foundation.
Abstract (Originalsprache)
Ground-penetrating radar 3D imaging holds significant value in geological exploration and cultural heritage preservation. However, traditional methods suffer from low resolution and high computational complexity, while existing deep learning models often feature complex architectures that are difficult to deploy on resource-constrained platforms. To address this, this paper proposes a lightweight 3D imaging network, 3D LiteUnet. By integrating an inverted residual module with a channel attention mechanism and introducing a multi-scale feature fusion strategy, it enhances reconstruction capabilities for multi-scale targets while reducing computational overhead. Experiments demonstrate the model’s superior performance on both simulation and measurement datasets: metrics including Dice score outperform existing methods, while offering advantages in parameter count, model size, and inference speed. Balancing imaging accuracy with deployment efficiency, this approach provides a novel solution for low-computational-power GPR 3D imaging.
Förderung
- Science and Technology Support Project for the Construction of Ordos National Innovation Demonstration Zone for the Sustainable Development Agenda
- Youth Fund of the Inner Mongolia Natural Science Foundation
Zitieren
APA
Yin, Y., Li, M., Pan, B., Du, Y., & Fan, C. (2026). Deep learning-based 3D imaging study of ground-penetrating radar. Physica Scripta, 101(9), 095003. https://doi.org/10.1088/1402-4896/ae4790
Vancouver
Yin Y, Li M, Pan B, Du Y, Fan C. Deep learning-based 3D imaging study of ground-penetrating radar. Physica Scripta. 2026;101(9):095003. doi:10.1088/1402-4896/ae4790
BibTeX
@article{Yin2026,
author = {Yin, Yongchao and Li, Minchao and Pan, Bingbing and Du, Yongxing and Fan, Changchun},
title = {Deep learning-based 3D imaging study of ground-penetrating radar},
journal = {Physica Scripta},
year = {2026},
volume = {101},
number = {9},
pages = {095003},
doi = {10.1088/1402-4896/ae4790}
}
EUVIMED