Unsupervised dual-branch attention network for multifrequency ground-penetrating radar data fusion: an application to Arctic sea ice exploration

Abstract

Ground-penetrating radar (GPR) is a nondestructive electromagnetic (EM) method used for subsurface target detection and stratigraphic imaging. The use of a single-frequency GPR profile is inherently limited by the trade-off between penetration depth and vertical resolution. For example, low-frequency antennas provide deeper penetration with lower resolution, whereas high-frequency antennas perform better in shallow areas but attenuate rapidly. Hence, multifrequency GPR data fusion can have complementary advantages. Conventional fusion methods usually depend on handcrafted feature extraction and fusion rules, whereas supervised deep learning-based methods are difficult to apply in polar field surveys due to the unavailability of absolute paired ground-truth labels. To address these limitations, this study proposes an unsupervised dual-branch attention network (UDBA-Net) for the fusion of 400/900 MHz multifrequency GPR data and applies it to Arctic multiyear sea ice exploration. The proposed network combines frequency-specific dual-branch encoders, a convolutional block attention module (CBAM)-like channel- or spatial-attention recalibration module, and an unsupervised hybrid loss comprising local intensity, structural similarity, and Laplacian-gradient constraints. The low- and high-frequency profiles are encoded separately to reduce early cross-band feature aliasing, whereas the attention module adaptively recalibrates the concatenated features based on channel contribution and spatial reflection-event salience. The main objective of the method is the structural and morphological enhancement, rather than preservation, of absolute electromagnetic reflection coefficients. Experimental analyses using controlled synthetic data and Arctic field GPR profiles confirm that UDBA-Net enhances relative structural contrast, improves reflection-event continuity, and integrates shallow, high-resolution textures with deeper, low-frequency interfaces under label-scarce conditions.

Keywords

Ground-penetrating radar / Multifrequency data fusion / Unsupervised learning / Dual-branch architecture / Arctic sea ice

References

[1] Bi WD, Zhao YH, Shen RQ, Li B, Hu SF, Ge SC. Multi-frequency GPR data fusion and its application in NDT. NDT E Int, 2020, 115 102289

[2] De Coster A, Lambot S (2016) Multi-frequency GPR data fusion. In: 2016 16th International Conference on Ground Penetrating Radar (GPR). IEEE, pp 1–6. https://doi.org/10.1109/ICGPR.2016.7572512

[3] Dinh K, Gucunski N, Duong TH. An algorithm for automatic localization and detection of rebars from GPR data of concrete bridge decks. Autom Constr, 2018, 89: 292-298

[4] Fang L, Yang F, Fang YJ, Nie JL. Multi-frequency GPR image fusion based on convolutional sparse representation to enhance road detection. J Imaging, 2026, 12: 52

[5] Guan ZL, Liu WL. Multi-frequency GPR data fusion through a joint sliding window and wavelet transform-weighting method for top-coal structure detection. Appl Sci, 2024, 14 2721

[6] Hu YW, Ge YS, Qi MM, Xu SH. Dual branch attention network for image super-resolution. Sci Rep, 2025, 15 29019

[7] Kwok R. Arctic sea ice thickness, volume, and multiyear ice coverage: losses and coupled variability (1958–2018). Environ Res Lett, 2018, 13 105005

[8] Lee D, Bae HS, Jang H (2025) Attribute-based interpretation of ground-penetrating radar data for improved sea ice thickness estimation. Lithosphere 2025(3):136. https://doi.org/10.2113/2025/lithosphere_2025_136

[9] Li H, Wu XJ. DenseFuse: a fusion approach to infrared and visible images. IEEE Trans Image Process, 2019, 28: 2614-2623

[10] Liu Z, Blasch E, Xue ZY, Zhao JY, Laganiere R, Wu W. Objective assessment of multiresolution image fusion algorithms for context enhancement in night vision: a comparative study. IEEE Trans Pattern Anal Mach Intell, 2012, 34: 94-109

[11] Lu GX, Zhao WK, Forte E, Tian G, Li Y, Pipan M. Multi-frequency and multi-attribute GPR data fusion based on 2-D wavelet transform. Measurement, 2020, 166 108243

[12] Luo X, Yang XP, Gong JB, Lan T (2024) A multi-frequency GPR data fusion technology based on end-to-end deep neural networks. In: IET International Radar Conference (IRC 2023). IEEE, pp 1271–1276. https://doi.org/10.1049/icp.2024.1269

[13] Ma JY, Yu W, Liang PW, Li C, Jiang JJ. FusionGAN: a generative adversarial network for infrared and visible image fusion. Inform Fusion, 2019, 48: 11-26

[14] Meier WN, Petty A, Hendricks S, Bliss A, Kaleschke L, Divine D et al (2025) NOAA Arctic report card 2025: sea ice. In: NOAA technical report OAR ARC 25–06, pp 1–9. https://doi.org/10.25923/mmxf-0r86

[15] Prabhakar KR, Srikar VS, Babu RV (2017) DeepFuse: a deep unsupervised approach for exposure fusion with extreme exposure image pairs. In: 2017 IEEE International Conference on Computer Vision (ICCV). IEEE, pp 4724–4732. https://doi.org/10.1109/ICCV.2017.505

[16] Ran JL, Zhang WM, Yu Y. Research progress of deep learning in sea ice prediction. Remote Sens, 2026, 18: 419

[17] Tang LF, Yuan JT, Zhang H, Jiang X, Ma J. PIAFusion: a progressive infrared and visible image fusion network based on illumination aware. Inform Fusion, 2022, 83–84: 79-92

[18] Wang JW, Jiang M, Kong J. MDAN: multilevel dual-branch attention network for infrared and visible image fusion. Opt Lasers Eng, 2024, 176 108042

[19] Woo S, Park J, Lee JY, Kweon IS (2018) CBAM: convolutional block attention module. In: Ferrari V et al (eds) Lecture notes in computer science, vol 11211. Springer, Cham, pp 3–19. https://doi.org/10.1007/978-3-030-01234-2_1

[20] Xiao JP, Liu LB (2015) Multi-frequency GPR signal fusion using forward and inverse S-transform for detecting railway subgrade defects. In: 2015 8th International Workshop on Advanced Ground Penetrating Radar (IWAGPR). IEEE, pp 1–4. https://doi.org/10.1109/IWAGPR.2015.7292680

[21] Xing JH, Song R, Zhen S, Jiang XD. Construction of the central Arctic sea ice structure and acoustic velocity model at the short-term ice station during N11 CHINARE. IEEE Trans Geosci Remote Sens, 2022, 60: 5906307

[22] Zhang B, Dai QW, Yin XB, Li ZW, Feng DS, Sun Y. An optimized choice of UCPML to truncate lattices with rotated staggered grid scheme for ground penetrating radar simulation. IEEE Trans Geosci Remote Sens, 2019, 57: 8695-8706

[23] Zhao WK, Lu GZ. A novel multifrequency GPR data fusion algorithm based on time-varying weighting strategy. IEEE Geosci Remote Sens Lett, 2022, 19: 3510104

[24] Zhao WK, Yuan L, Forte E, Lu GZ, Tian G, Pipan M. Multi-frequency GPR data fusion with genetic algorithms for archaeological prospection. Remote Sens, 2021, 13 2804

Funding

National Natural Science Foundation of China(42076224)

National Key R&D Program of China(2021YFC2801202)

Rights & permissions

The Author(s)