Ground-penetrating radar (GPR) data are often contaminated by random noise in the actual engineering inspection.The noise in data will reduce the signal-to-noise ratio and resolution of the data,adversely affecting the subsequent inversion and interpretation.Accordingly,this study proposed neural network-based denoising for GPR data.First,a multi-layer neural network model was constructed to integrate the data corrupted by white Gaussian noise into the noise-free data.Then,the corrupted data and their corresponding noise patches were built as training data.The weights of neurons in every layer of the model e updated using a back-propagation algorithm to minimize the model training loss.Finally,the two synthetic data and the measured radar data were input to the trained model,and the model's output was calculated using the noise characteristic weights acquired from the training.Compared with the curvelet transform,the numerical simulation test results verify the effectiveness and robustness of the method proposed in this study.Moreover,the proposed method can suppress the noise more thoroughly in areas with complex structures and weak amplitudes,and show effective signals more clearly.
Dai Q W, Wu K J, Zhang B. A study of application of short-time Fourier transform to GPR data interpretation[J]. Geophysical and Geochemical Exploration, 2016, 40(6):1227-1231.
Zhang S W, Wu R X, Han Z A, et al. The application of bilateral filtering to denoise processing of ground penetrating radar data[J]. Geophysical and Geochemical Exploration, 2021, 45(2):496-501.
Dai Q W, Cheng Q Y, Feng D S. Low signal-noise ratio GPR signal denoising based on FastICA[J]. Computing Techniques for Geophysical and Geochemical Exploration, 2017, 39(6):727-735.
Su Z G, Liao J J, Qian D H. The interference signal in GPR survey data and its filtering[J]. Geophysical and Geochemical Exploration, 2011, 35(3):427-430.
[5]
周威帆. 基于f-x域模态分解的探地雷达数据处理与解释[D]. 长春: 吉林大学, 2020.
[5]
Zhou W F. Data processing and interpretation of GPR based on mode decomposition in f-x domain[D]. Changchun: Jilin University, 2020.
Wang C, Shen F M. Study of wavelet transform in ground penetration radar weak signal denoising[J]. Geophysical and Geochemical Exploration, 2015, 39(2):421-424.
Li J H, He Z X, Yang J, et al. Scale and rotation statistic-based self-adaptive function for ground penetrating radar denoising in curvelet domain[J]. Acta. Phys. Sin., 2019, 68(9):74-83.
Wu X L, Yan F, Zhen R, et al. Research on adaptive clutter suppression for ground penetrating radar based on wavelet transform and K-SVD[J]. Journal of Hebei University of Science and Technology, 2021, 42(2):111-118.
Zhang D H, Qin H. Tunnel primary support detection using ground penetrating radar and deep learning[J]. Modern Tunnelling Technology, 2020, 57(S1):174-178.
[12]
Han D, Tao L, Leng J B, et al. GCN:GPU-based cube CNN framework for hyperspectral image classification[C]// International Conference on Parallel Processing, 2017, 46:41-49.
Xu X J, Gou Y Y, Yang F. Research on early warning Model of roadbed diseases under urban roads based on GPR and probabilistic neural network[J]. Science Technology and Engineering, 2017, 17(17):118-124.
Lyu Y B, Zhao J W, Cao F L. Image denoising algorithm based on composite convolutional neural network[J]. Pattern Recognition and Artificial Intelligence, 2017, 30(2):97-105.
Li X C, Ye J W, Li G, et al. Elman neural network dynamic prediction based on landslide monitoring data[J]. Coal Geology & Exploration, 2018, 46(3):113-120,126.
[17]
Silver D, Huang A, Maddision C J, et al. Mastering the game of go with deep neural networks and tree search[J]. Nature, 2016, 529(7587):484-489.
doi: 10.1038/nature16961
[18]
Krizhevsky A, Sutskever I, Hinton G E. ImageNet classification with deep convolutional neural networks[J]. Communication ACM, 2017, 60(6):84-90.
doi: 10.1145/3065386
[19]
Jain V, Seung H S. Natural image denoising with convolutional networks[J]. Advances in Neural Information Processing Systems(NIPS), 2008, 21:769-776.
[20]
Lecun Y, Bottou L, Bengio Y, et al. Gradient-based learning applied to document recognition[J]. Proceedings of the IEEE, 1998, 86(11):2278-2324.
doi: 10.1109/5.726791
[21]
Chen S, Abhinav S, Saurabh S, et al. Revisiting unreasonable effectiveness of data in deep learning era[C]// 2017 IEEE International Conference on Computer Vision (ICCV), 2017:843-852.
Zhou F Y, Jin L P, Dong J. Review of convolutional neural network[J]. Chinese Journal of Computers, 2017, 40(6):1229-1251.
[23]
Hornik K, Stinchcombe M, White H. Multilayer feedforward networks are universal approximators[J]. Neural Networks, 1989, 2(5):359-366.
doi: 10.1016/0893-6080(89)90020-8
[24]
Burger H C, Schuler C J, Harmeling S, et al. Image denoising:Can plain neural networks compete with BM3D?[C]// Computer Vision and Pattern Recognition, 2012:2392-2399.
[25]
杨光照. 基于探地雷达的煤岩界面识别技术研究[D]. 徐州: 中国矿业大学, 2019.
[25]
Yang G Z. Research on coal-rock interface recognition technology based on ground penetrating radar[D]. Xuzhou: China University of Mining and Technology, 2019.
Feng D S, Yang Z L. Automatic recognition of ground penetrating radar image of tunnel lining structure based on deep learning[J]. Progress in Geophysics, 2020, 35(4):1552-1556.
[27]
季银涛. 基于深度学习的探地雷达图像介电常数反演研究[D]. 济南: 山东大学, 2021.
[27]
Ji Y T. Deep learning based ground penetrating radar image permittivity inversion research[D]. Jinan: Shandong University, 2021.
[28]
Sergey I, Christian S. Batch normalization:Accelerating deep network training by reducing internal covariate shift[C]// International Conference on Machine Learning,PMLR, 2015:448-456.
