Neural Network Methods for Recognizing Underground Objects in the Data of a Mobile Wheeled Robot's Ground-Penetrating Radar

Currently, ground-penetrating radar (GPR) is often used to solve problems related to searching for and recognizing underground objects. This allows for obtaining rich information about the studied area at various depths, but manual interpretation of radargrams is labor-intensive. This paper proposes the GeoRTr method for automatically detecting underground objects of interest. It is based on an original neural network architecture with a GeoAttention mechanism. The network's input is A-scans from a GPR mounted on a wheeled robot moving in automatic or semi-automatic mode. The proposed neural network model accounts for spatial relationships between A-scans even when the robot's trajectory deviates from a straight line. Using a mobile wheeled robot and a GPR with an antenna mounted at a low altitude above the surface being analyzed, a unique dataset was collected and labeled for recognizing several categories of subsurface objects. Experiments with the developed neural network model on the collected dataset demonstrated a classification F1-score of over 40%, surpassing comparable methods. A comparative analysis of various model architecture variations identified two configurations for different ground-penetrating radar object search conditions.

1. Wu, M., Liu, Q., & Ouyang, S. (2025). Two-stage GPR image inversion method based on multi-scale dilated convolution and hybrid attention gate. Remote Sensing, 17(2), Article 322. https://doi.org/10.3390/rs17020322

2. Lee, K., & Feng, C. (2025). GPR-Former: Context-aware moisture detection. In ISARC. Proceedings of the International Symposium on Automation and Robotics in Construction (pp. 1–8). IAARC Publications. https://doi.org/10.22260/isarc2025/0002

3. Jeon, M., Lee, H., & Lee, J. (2026). Efficient and robust temporal modeling for LGPR odometry with Mamba. IEEE Access, 14, 38775–38792. https://doi.org/10.1109/ACCESS.2026.36

4. Tegegne, M. T., Yang, R., Huang, S., Xiao, H., Yang, Z., Zhang, Q., Liu, Y., & Qi, W. (2026). Automated railway subgrade defect detection using GPR and an enhanced YOLOv11 framework. Transportation Geotechnics, 62, Article 102105. https://doi.org/10.1016/j.trgeo.2026.1

5. Lei, W., Yu, S., & Zhang, T. (2026). HFL-YOLOv8: A hyperbolic feature-enhanced lightweight network for object detection in ground penetrating radar images. Applied Soft Computing, 188, Article 114403. https://doi.org/10.1016/j.asoc.2025.114403

6. Zheng, P., Zhang, A., Shi, Z., Wang, S., Ma, Y., & Liu, Z. (2025). TLAD-YOLO: Lightweight network for intelligent detection of railway tunnel lining anomalies using ground penetrating radar. Journal of Applied Geophysics, 241, Article 105869. https://doi.org/10.1016/j.jappgeo.2025.105869

7. Cao, S., Lu, C., Wang, X., Zhang, P., Jin, G., & Cai, W. (2026). ME-YOLO: A novel real-time detection network for pavement interlayer distress using ground-penetrating radar. Journal of Applied Geophysics, 245, Article 106057. https://doi.org/10.1016/j.jappgeo.2025.106

8. Zhang, W., Liu, Z., Yang, Q., & Qiu, X. (2026). FLS-YOLO: A multi-scale subsurface defect detection network via heterogeneous grouping and spatial reconstruction. Signal, Image and Video Processing, 20(5), Article 262. https://doi.org/10.1007/s11760-026-05352-z

9. Kang, M., Kaji, S., Lee, S.-Y., Kim, T., Ryu, H.-H., & Choi, S. (2025). A novel shapeaware topological representation for GPR data with DNN integration (arXiv:2506.06311). arXiv. https://doi.org/10.48550/arXiv.2506.06311

10. Arvas, M. M., & Çınar, A. (2026). Detection of underground objects from GPR data using a lightweight YOLO-based approach. Scientific Reports. https://doi.org/10.1038/s41598-0-59135-0

11. Lv, H., Zhang, Y., Dai, J., Wu, H., Wang, J., & Wang, D. (2025). Intelligent recognition of GPR road hidden defect images based on feature fusion and attention mechanism. IEEE Transactions on Geoscience and Remote Sensing, 63, 1–17.

12. Mojahid, A., El Ouai, D., El Amraoui, K., El-Hami, K., Aitbenamer, H., Verrelst, J., & Barone, P. M. (2025). Lightweight CNN model for automatic detection and depth estimation of subsurface voids using GPR B-scan data. Natural Hazards Research, 5(2), 432–446. https://doi.org/10.1016/j.nhres.2025.02.001

13. Wang, C., Guan, Y., Chi, M., Shen, F., Yu, Z., Chen, Q., & Chen, C. (2025). GPR-TSBiNet: An information gradient enrichment model for GPR B-scan small target detection. Sensors, (7), Article 2223. https://doi.org/10.3390/s25072223

14. Wang, H., et al. (2025). Attention-module-guided time-lapse leakage plume imaging driven by LeakInv-CUNet GPR inversion framework. IEEE Access, 13, 122514–122529.

15. Yang, S., Guo, L., Yang, Y., & Ye, H. (2026). Enhanced Pix2pixGAN with spatial-channel attention for underground medium inversion from GPR. Remote Sensing, 18(3), Article 448. https://doi.org/10.3390/rs18030448