Fast Efficiency Googlenet using GPR Signal Processing

Abstract

Ground Penetrating Radar (GPR) is a geophysical locating method that uses radio waves to capture images under the ground\'s surface in a least invasive way. Non-destructive testing relies heavily on the localisation and reconstruction of subsurface targets, as well as the calculation of object position and form utilising Ground Penetration Radar (GPR). In this design, the raw GPR data must be pre-processed. This data is then denoised using the DWT transform, and time- frequency data is processed with Fast efficient Googlenet. This approach applies the idea of global relative optimality from fast efficient deep learning to overcome the limitations of traditional algorithms that concentrate on extracting absolute features and therefore improving accuracy.

Introduction

1. Ground Penetrating Radar (GPR) Technology:

• GPR is a non-destructive geophysical method used to detect buried metal pipes by emitting high-frequency radar waves.

• When these waves hit metal pipes, strong reflections are captured, helping determine the position, depth, and shape of the pipes.

• GPR is advantageous over traditional methods (e.g., excavation) due to its speed, safety, and efficiency.

2. Challenges in Metal Pipe Detection:

• Difficulties arise from complex soil conditions, pipe material, depth, and noise/clutter in the data.

• Metal pipes may create misleading signals or clutter, complicating interpretation.

• High soil moisture and corroded pipes can attenuate radar signals, reducing effectiveness.

3. Application and Importance:

• Accurate detection is vital for urban infrastructure, utility maintenance, and construction safety.

• Helps avoid accidental pipe damage, reduce costs, and prevent environmental hazards.

4. Use of Deep Learning for GPR Data Interpretation:

• Manual analysis of GPR B-scan images (which display radar reflections) is slow and subjective.

• Deep learning, especially Convolutional Neural Networks (CNNs) like Dense GoogLeNet, is used to automatically detect and classify hyperbolic reflections of underground pipelines.

5. Methodology:

• The study uses 100 real GPR B-scan images of 512×512 resolution.

• A deep CNN architecture processes GPR images, enhancing features, reducing noise, and identifying hyperbolic patterns caused by pipelines.

• Techniques like ReLU activation, pooling, batch normalization, and dropout are employed to improve model performance.

• The CNN model can distinguish between threat/non-threat profiles and achieves high accuracy in detecting underground utilities.

6. Literature Review Highlights:

• Multiple researchers have proposed methods such as MigrationNet, YOLOv3, and 3D visualizations for pipeline detection using GPR.

• Techniques like the Hough Transform, Radon Transform, and time-frequency feature extraction have been used to improve detection accuracy.

• GPR is widely used in archaeology, explosive detection, and infrastructure monitoring.

Conclusion

This study shows applications for detecting and localising subsurface pipelines. This study created a labor-saving way for boosting training data, mostly via the analysis of GPR findings using deep neural networks, and shown that the system can locate subsurface pipes with good accuracy when applied on real roads. Our suggested strategy offers various benefits over standard methods for boosting training data. First, it can create synthetic pictures that are comparable to genuine photos, considerably increasing the training dataset\'s variety and representativeness. Second, it can automatically annotate the created photos, saving time and reducing human error throughout the annotation process. Third, it may iteratively create new pictures from old ones, increasing the quantity and quality of the training dataset.

References

[1] Iftimie, N.; Savin, A.; Steigmann, R.; Dobrescu, G.S. Underground Pipeline Identification into a Non- Destructive Case Study Based on Ground-Penetrating Radar Imaging. Remote Sens. 2021, 13, 3494. https://doi.org/10.3390/rs13173494 [2] Martínez, Francisco & Solla, Mercedes & Puente, Iván & Arias, Pedro. (2017). Efficient GPR data acquisition to detect underground pipes. NDT & E International. 91. 10.1016/j.ndteint.2017.06.002. [3] Feng, Jinglun & Yang, Liang & Wang, Haiyan & Tian, Yingli & Xiao, Jizhong. (2021). Subsurface Pipes Detection Using DNN-based Back Projection on GPR Data. 266-275. 10.1109/WACV48630.2021.00031. [4] Hoarau, Q & Ginolhac, Guillaume & Atto, A.M. & Nicolas, Jean-Marie & Ovarlez, Jean-Philippe. (2016). Robust Adaptive Detection of Buried Pipes using GPR. 10.1109/EUSIPCO.2016.7760305. [5] W. Sun, Q. Xu, H. Zhang and Z. Yao, \"Research on Detection and Visualization of Underground Pipelines,\" 2012 2nd International Conference on Remote Sensing, Environment and Transportation Engineering, Nanjing, China, 2012, pp. 1-4, doi: 10.1109/RSETE.2012.6260692. [6] Hai Liu, Yunpeng Yue, Chao Liu, B.F. Spencer, Jie Cui, Automatic recognition and localization of underground pipelines in GPR B-scans using a deep learning model, Tunnelling and Underground Space Technology, Volume 134, 2023, 104861, ISSN 0886-7798, [7] https://doi.org/10.1016/j.tust.2022.104861. [8] Bai, Xu, Yu Yang, Zhitao Wen, Shouming Wei, Jiayan Zhang, Jinlong Liu, Hongrui Li, Haoxiang Tian, and Guanting Liu. 2023. \"3D-GPR-RM: A Method for Underground Pipeline Recognition Using 3- Dimensional GPR Images\" Applied Sciences 13, no. 13: 7540. https://doi.org/10.3390/app13137540 [9] Yang Su, Jun Wang, Danqi Li, Xiangyu Wang, Lei Hu, Yuan Yao, Yuanxin Kang, End-to-end deep learning model for underground utilities localization using GPR, Automation in Construction, [10] Volume 149, 2023, 104776, ISSN 0926-5805, https://doi.org/10.1016/j.autcon.2023.104776. [11] G. Yang, D. Yuan, T. Xu and B. Li, \"An Adaptive Clutter-Immune Method for Pipeline Detection With GPR,\" in IEEE Sensors Journal, vol. 23, no. 19, pp. 22984-22997, 1 Oct.1, 2023, doi: 10.1109/JSEN.2023.3305681. [12] Q. Xu, Z. Wang and X. Li, \"Depth and Radius Joint Estimation for Underground Pipeline Using GPR,\" in IEEE Sensors Journal, vol. 23, no. 13, pp. 14628-14639, 1 July1, 2023, doi: 10.1109/JSEN.2023.3280177. [13] Lu, Q.; Pu, J.; Liu, Z. Feature extraction and automatic material classification of underground objects from ground penetrating radar data. J. Electr. Comput. Eng. 2014, 14, 28. [14] Mukhopadhyay, P.; Chaudhuri, B.B. A survey of Hough Transform. Pattern Recognit. 2015, 48, 993– 1010. [15] W.; Cui, X.; Guo, L.; Chen, J.; Chen, X.; Cao, X. Tree Root Automatic Recognition in Ground Penetrating Radar Profiles Based on Randomized Hough Transform. Remote Sens. 2016, 8, 430. [16] Windsor, C.G.; Capineri, L.; Falorni, P. A Data Pair-Labeled Generalized Hough Transform for Radar Location of Buried Objects. IEEE Geosci. Remote Sens. Lett. 2013, 11, 124–127. [17] Dell’Acqua, A.; Sarti, A.; Tubaro, S.; Zanzi, L. Detection of linear objects in GPR data. Signal Process. 2004, 84, 785–799. [18] L. Chenglong et.al. Advances in automatic identification of road subsurface distress using ground penetrating radar: State of the art and future trends. Automation in Construction, (2024). 158. 105185. 10.1016/j.autcon.2023.105185. [19] L. Fanruo et.al., 3D ground penetrating radar road underground target identification algorithm using time-frequency statistical features of data. NDT & E International (2023). 137. 102860. 10.1016/j.ndteint.2023.102860. [20] K. Toshinori et.al., Estimation of buried pipe depth in an artificial soil tank using ground-penetrating radar and moisture sensor. Journal of Applied Geophysics,(2024) . 220. 105283. 10.1016/j.jappgeo.2023.105283. [21] M. Nasri et.al.; Recognition of different utility pipes size of ground penetrating radar images at different penetration depth. AIP Conf. Proc. 8 February 2024; 2898 (1): 030067. https://doi.org/10.1063/5.0194126

Copyright

Copyright © 2025 Priya B. Shinde, Aziz Binnaser, Gaju S. Chavan , Shital N. Katkade. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.