DR-GAN:一種無(wú)監(jiān)督學(xué)習(xí)的探地雷達(dá)雜波抑制方法
doi: 10.11999/JEIT221072
-
1.
中南大學(xué)計(jì)算機(jī)學(xué)院 長(zhǎng)沙 410083
-
2.
中國(guó)電波傳播研究所 青島 266107
DR-GAN: An Unsupervised Learning Approach to Clutter Suppression for Ground Penetrating Radar
-
1.
School of Computer Science, Central South University, Changsha 410083, China
-
2.
China Research Institute of Radiowave Propagation, Qingdao 266107, China
-
摘要: 探地雷達(dá)(GPR)是一種基于電磁波的地下無(wú)損探測(cè)技術(shù),廣泛應(yīng)用于市政工程、交通、軍事等領(lǐng)域。在數(shù)據(jù)采集過(guò)程中,由于發(fā)射天線(xiàn)和接收天線(xiàn)之間的耦合、起伏地面的散射以及地下隨機(jī)媒質(zhì)的復(fù)雜性等原因,采集得到的GPR B-scan回波中通常存在雜波,雜波嚴(yán)重影響了地下目標(biāo)的檢測(cè)和特征提取。該文提出一種用于GPR B-scan圖像雜波抑制的解糾纏表示生成對(duì)抗網(wǎng)絡(luò)(DR-GAN),設(shè)計(jì)了目標(biāo)特征編碼器和雜波特征編碼器用來(lái)提取GPR B-scan圖像中的目標(biāo)特征和雜波特征,設(shè)計(jì)了雜波抑制生成器用來(lái)獲取雜波抑制后的GPR B-scan圖像。與現(xiàn)有的基于監(jiān)督學(xué)習(xí)的GPR雜波抑制方法相比,該方法在網(wǎng)絡(luò)訓(xùn)練時(shí)不需要成對(duì)的匹配數(shù)據(jù),可以更好地應(yīng)用于實(shí)測(cè)GPR圖像的雜波抑制。在仿真和實(shí)測(cè)GPR數(shù)據(jù)上的實(shí)驗(yàn)結(jié)果表明,DR-GAN這一無(wú)監(jiān)督學(xué)習(xí)網(wǎng)絡(luò)具有更好的雜波抑制性能。對(duì)石英砂中埋設(shè)的鋼筋進(jìn)行數(shù)據(jù)采集,運(yùn)用DR-GAN對(duì)含雜波的實(shí)測(cè)數(shù)據(jù)進(jìn)行處理,處理結(jié)果的改善系數(shù)(IF)指標(biāo)較現(xiàn)有的魯棒非負(fù)矩陣分解(RNMF)方法提高了17.85 dB。
-
關(guān)鍵詞:
- 探地雷達(dá) /
- 雜波抑制 /
- 無(wú)監(jiān)督學(xué)習(xí) /
- 解糾纏表示 /
- 生成對(duì)抗網(wǎng)絡(luò)
Abstract: Ground Penetrating Radar (GPR) is an underground nondestructive detection technology based on electromagnetic wave, which is widely used in municipal engineering, transportation, military and other fields. In the process of data acquisition, due to the coupling between transmitting antenna and receiving antenna, scattering from undulating ground and the complexity of underground random media, there is usually clutter in the GPR B-scan, which affects seriously the detection and feature extraction of underground targets. A Disentanglement Representation Generative Adversarial network (DR-GAN) for clutter suppression in GPR B-scan images is proposed. A target feature encoder and a clutter feature encoder are designed to extract target features and clutter features in GPR B-scan images. A clutter suppression generator is designed to obtain the GPR B-scan image after clutter suppression. Compared with the existing GPR clutter suppression methods based on supervised learning, the proposed method does not need pairwise matching data during network training, and can be better applied to the clutter suppression of measured GPR images. Experimental results on simulated and measured GPR data show that DR-GAN is an unsupervised learning network with better clutter suppression performance. The data of reinforcement embedded in quartz sand are collected, and the measured data containing clutter are processed by DR-GAN. The Improvement Factor (IF) index of the processing results is 17.85 dB higher than that of the existing Robust Nonnegative Matrix Factorization (RNMF) method. -
表 1 仿真場(chǎng)景參數(shù)
模型參數(shù) 掃描場(chǎng)景參數(shù) 模型尺寸 1.8 m×0.002 m×0.45 m 發(fā)射源波形 瑞克子波 單元格大小 0.002 m×0.002 m×0.002 m 發(fā)射中心頻率 2 GHz 土壤沙子重量百分比 50% 仿真時(shí)窗 10 ns 土壤粘土重量百分比 50% 發(fā)射天線(xiàn)起點(diǎn) (0.1 m,0.002 m,0.4 m) 土壤的容重 2.0 g/cm3 接收天線(xiàn)起點(diǎn) (0.2 m,0.002 m,0.4 m) 土壤的沙粒密度 2.66 g/cm3 每次掃描天線(xiàn)移動(dòng)距離 0.01 m 土壤體積含水率范圍 0.001~0.150 目標(biāo)形狀 圓柱、方柱 土壤材料種類(lèi) 50種 圓柱半徑 0.03~0.05 m 土壤厚度 0.4 m 方柱邊長(zhǎng) 0.04~0.06 m 土壤表面高度起伏范圍 0.38~0.41 m 目標(biāo)高度 0.2~0.3 m 目標(biāo)水平位置 0.35~1.45 m 下載: 導(dǎo)出CSV
表 2 各種雜波抑制算法的平均PSNR(dB)/平均SSIM
目標(biāo)類(lèi)型 MS SVD NMF RPCA RNMF RAE 本文DR-GAN PVC圓柱體 0.85/0.014 22.38/0.365 22.23/0.305 23.54/0.259 22.72/0.205 22.56/0.154 36.66/0.965 空洞圓柱體 6.62/0.060 22.96/0.420 22.96/0.398 25.94/0.470 24.35/0.338 22.99/0.363 42.73/0.989 金屬圓柱體 12.58/0.126 25.26/0.588 25.18/0.562 27.59/0.612 25.86/0.471 25.92/0.432 44.70/0.992 PVC方柱體 6.41/0.021 23.09/0.357 24.06/0.365 27.11/0.437 25.84/0.351 25.22/0.287 43.65/0.980 空洞方柱體 12.45/0.072 25.79/0.535 25.38/0.494 28.78/0.597 26.43/0.444 25.80/0.369 47.92/0.993 金屬方柱體 16.97/0.181 25.97/0.538 25.39/0.504 29.15/0.652 27.45/0.538 26.72/0.458 48.00/0.993 下載: 導(dǎo)出CSV
表 3 各種雜波抑制方法的平均用時(shí)(s)
MS SVD NMF RPCA RNMF RAE DR-GAN(GPU) 時(shí)間 0.002 9 0.010 2 0.013 6 2.083 9 6.063 6 3.872 6 0.103 0 下載: 導(dǎo)出CSV
表 4 實(shí)測(cè)數(shù)據(jù)雜波抑制的平均IF(dB)
目標(biāo)類(lèi)型 MS SVD NMF RPCA RNMF RAE 本文DR-GAN 空心PVC管 10.49 24.19 24.59 20.68 24.74 12.01 41.34 空心塑料瓶 13.26 24.50 23.82 23.18 24.67 18.40 45.79 鋼筋 13.19 26.07 25.96 26.28 28.16 26.65 46.01 下載: 導(dǎo)出CSV
-
[1] 喬爾, 雷文太, 童孝忠, 周腸, 等譯. 探地雷達(dá)理論與應(yīng)用[M]. 北京: 電子工業(yè)出版社, 2011.JOL H M, LEI Wentai, TONG Xiaozhong, ZHOU Yang, et al. translation. Ground Penetrating Radar: Theory and Applications[M]. Beijing: Publishing House of Electronics Industry, 2011. [2] 劉瀾波, 錢(qián)榮毅. 探地雷達(dá): 淺表地球物理科學(xué)技術(shù)中的重要工具[J]. 地球物理學(xué)報(bào), 2015, 58(8): 2606–2617. doi: 10.6038/cjg20150802LIU Lanbo and QIAN Rongyi. Ground penetrating radar: A critical tool in near-surface geophysics[J]. Chinese Journal of Geophysics, 2015, 58(8): 2606–2617. doi: 10.6038/cjg20150802 [3] TONG Zheng, YUAN Dongdong, GAO Jie, et al. Pavement-distress detection using ground-penetrating radar and network in networks[J]. Construction and Building Materials, 2020, 233: 117352. doi: 10.1016/j.conbuildmat.2019.117352 [4] 劉海, 黃肇剛, 岳云鵬, 等. 地下管線(xiàn)滲漏環(huán)境下探地雷達(dá)信號(hào)特征分析[J]. 電子與信息學(xué)報(bào), 2022, 44(4): 1257–1264. doi: 10.11999/JEIT211213LIU Hai, HUANG Zhaogang, YUE Yunpeng, et al. Characteristics analysis of ground penetrating radar signals for groundwater pipe leakage environment[J]. Journal of Electronics &Information Technology, 2022, 44(4): 1257–1264. doi: 10.11999/JEIT211213 [5] SOLIMENE R, CUCCARO A, DELL' AVERSANO A, et al. Ground clutter removal in GPR surveys[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2014, 7(3): 792–798. doi: 10.1109/JSTARS.2013.2287016 [6] ABUJARAD F, NADIM G, and OMAR A. Clutter reduction and detection of landmine objects in ground penetrating radar data using Singular Value Decomposition (SVD)[C]. The 3rd International Workshop on Advanced Ground Penetrating Radar, Delft, Netherlands, 2005: 37–41. [7] CHEN Gaoxiang, FU Liyun, CHEN Kanfu, et al. Adaptive ground clutter reduction in ground-penetrating radar data based on principal component analysis[J]. IEEE Transactions on Geoscience and Remote Sensing, 2019, 57(6): 3271–3282. doi: 10.1109/TGRS.2018.2882912 [8] ABUJARAD F and OMAR A. Comparison of independent component analysis (ICA) algorithms for GPR detection of non-metallic land mines[J]. SPIE, 2006, 6365: 636516. [9] KUMLU D and ERER I. Performance evaluation of NMF methods with different divergence metrics for landmine detection in GPR[J]. SPIE, 2018, 10794: 107940I. [10] TEMLIOGLU E and ERER I. Clutter removal in ground-penetrating radar images using morphological component analysis[J]. IEEE Geoscience and Remote Sensing Letters, 2016, 13(12): 1802–1806. doi: 10.1109/LGRS.2016.2612582 [11] KUMLU D and ERER I. Improved clutter removal in GPR by robust nonnegative matrix factorization[J]. IEEE Geoscience and Remote Sensing Letters, 2020, 17(6): 958–962. doi: 10.1109/LGRS.2019.2937749 [12] SONG Xiaosong, XIANG Deliang, ZHOU Kai, et al. Improving RPCA-based clutter suppression in GPR detection of antipersonnel mines[J]. IEEE Geoscience and Remote Sensing Letters, 2017, 14(8): 1338–1342. doi: 10.1109/LGRS.2017.2711251 [13] NI Zhikang, YE Shengbo, SHI Cheng, et al. Clutter suppression in GPR B-scan images using robust autoencoder[J]. IEEE Geoscience and Remote Sensing Letters, 2022, 19: 3500705. doi: 10.1109/LGRS.2020.3026007 [14] ZHOU Huilin, WANG Yi, LIU Qiegen, et al. RNMF-guided deep network for signal separation of GPR without labeled data[J]. IEEE Geoscience and Remote Sensing Letters, 2022, 19: 3507705. doi: 10.1109/LGRS.2021.3099161 [15] GENG Jianrong, HE Juan, YE Hongxia, et al. A clutter suppression method based on LSTM network for ground penetrating radar[J]. Applied Sciences, 2022, 12(13): 6457. doi: 10.3390/app12136457 [16] TEMLIOGLU E and ERER I. A novel convolutional autoencoder-based clutter removal method for buried threat detection in ground-penetrating radar[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 5103313. doi: 10.1109/TGRS.2021.3098122 [17] NI Zhikang, SHI Cheng, PAN Jun, et al. Declutter-GAN: GPR B-scan data clutter removal using conditional generative adversarial nets[J]. IEEE Geoscience and Remote Sensing Letters, 2022, 19: 4023105. doi: 10.1109/LGRS.2022.3159788 [18] HUANG Yongqiang, XIA Wenjun, LU Zexin, et al. Noise-powered disentangled representation for unsupervised speckle reduction of optical coherence tomography images[J]. IEEE Transactions on Medical Imaging, 2021, 40(10): 2600–2614. doi: 10.1109/TMI.2020.3045207 [19] LEE H Y, TSENG H Y, HUANG Jiabin, et al. Diverse image-to-image translation via disentangled representations[C]. The 15th European Conference on Computer Vision, Munich, Germany, 2018: 36–52. [20] ABROL V, SHARMA P, and PATRA A. Improving generative modelling in VAEs using multimodal prior[J]. IEEE Transactions on Multimedia, 2021, 23: 2153–2161. doi: 10.1109/TMM.2020.3008053 [21] ISOLA P, ZHU Junyan, ZHOU Tinghui, et al. Image-to-image translation with conditional adversarial networks[C]. 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, USA, 2017: 5967–5976. [22] WARREN C, GIANNOPOULOS A, and GIANNAKIS I. gprMax: Open source software to simulate electromagnetic wave propagation for Ground Penetrating Radar[J]. Computer Physics Communications, 2016, 209: 163–170. doi: 10.1016/j.cpc.2016.08.020 -