Comptes Rendus
Article de recherche
Multi-label classification with deep learning techniques applied to the B-Scan images of GPR
[Classification multi-label par des techniques d’apprentissage profond appliquées aux images B-Scan de radars à sondage de sol]
Comptes Rendus. Physique, Online first (2024), pp. 1-16.

Les radars à pénétration de sol (GPR) sont aujourd’hui largement utilisés pour la détection d’objets enterrés dans des domaines tels que la géologie, l’archéologie et le génie civil. Il présente l’avantage de permettre une détection par une technique non destructive. Le principe du GPR dans le domaine temporel consiste à émettre des impulsions électromagnétiques dans le sol, celles-ci étant ensuite diffractées par les cibles à détecter. Une seule trace de signal GPR capturée à une position du radar est un signal 1D appelé Ascan. Un ensemble de formes d’ondes radar Ascan capturées à un certain nombre de positions consécutives différentes le long d’une direction particulière formera une image 2D appelée B-scan dans le cas d’un déplacement rectiligne. Elles montrent des formes de réponse de type hyperbolique et leur analyse donne de nombreuses caractéristiques. Par exemple, dans le cas de canalisations enterrées, un traitement spécifique permet de connaître leur diamètre, leur nature ainsi que les caractéristiques électriques du sol. Cependant, ces approches nécessitent souvent un post-traitement complexe du Bscan, ce qui peut être chronophage et rend donc difficile la caractérisation en temps réel au détriment de ces méthodes. Avec l’émergence des réseaux neuronaux profonds et avec une phase d’apprentissage sur un grand nombre de Bscan, il devient possible d’extraire presque instantanément les caractéristiques des données radar GPR. Dans cette étude, un modèle de classification multi-label (MLC) basé sur l’apprentissage par transfert et l’augmentation des données a été développé pour générer des éléments d’information multiples sur la même image et réaliser la classification. Trois modèles d’apprentissage profond : VGG-16, ResNet-50 et CNN adapté ont été utilisés comme modèles pré-entraînés pour l’apprentissage par transfert. Les réseaux ont été formés sur un ensemble de données synthétiques créé dans cette étude et évalués sur un ensemble de mesures de performance.

The ground penetrating radars (GPR) are now widely used for the detection of buried objects in areas such as: geology, archaeology and civil engineering. It has the advantage of allowing detection by a non-destructive technique. The principle for time domain GPR consists in emitting electromagnetic pulses in the ground, these one are then diffracted by the targets to be detected. A single GPR signal trace captured at a position of the radar is a 1D signal called Ascan. A set of Ascan radar waveforms captured at a certain number of different consecutive positions along a particular direction will form a 2D image called B-scan. They show response shapes of hyperbolic type and their analysis give many characteristics. For example, in the case of buried pipes, a specific processing allows to find their diameter, their nature as well as the electrical characteristics of the ground. However, these approaches often require complex post-processing of the Bscan, which can be time-consuming and therefore makes it difficult to perform real-time characterization at the expense of such methods. With the emergence of deep neural networks and with a learning phase on a large number of Bscan, it becomes possible to extract almost instantaneously the characteristics of GPR radar data. In this study, a multi-label classification (MLC) model based on transfer learning and data augmentation has been developed to generate multiple information elements on the same image and to realize classification. Three deep learning models: VGG-16, ResNet-50 and adapted CNN were used as pre-trained models for transfer learning. The networks were trained on a synthetic dataset created in this study and evaluated on a set of performance metrics.

Reçu le :
Révisé le :
Accepté le :
Première publication :
DOI : 10.5802/crphys.193
Keywords: Ground Penetrating Radar, Image processing, Detection of Buried objects, Deep learning
Mot clés : Radar à sondage de sol, traitement d’images, détection d’objets enfouis, apprentissage profond
Note : This article follows the URSI-France workshop held on 21 and 22 March 2023 at Paris-Saclay.

Soukayna El Karakhi 1 ; Alain Reineix 1 ; Christophe Guiffaut 1

1 University of Limoges, XLIM Institute, 123 Av. Albert Thomas, 87000 Limoges, France
Licence : CC-BY 4.0
Droits d'auteur : Les auteurs conservent leurs droits
@article{CRPHYS_2024__25_S1_A12_0,
     author = {Soukayna El Karakhi and Alain Reineix and Christophe Guiffaut},
     title = {Multi-label classification with deep learning techniques applied to the {B-Scan} images of {GPR}},
     journal = {Comptes Rendus. Physique},
     publisher = {Acad\'emie des sciences, Paris},
     year = {2024},
     doi = {10.5802/crphys.193},
     language = {en},
     note = {Online first},
}
TY  - JOUR
AU  - Soukayna El Karakhi
AU  - Alain Reineix
AU  - Christophe Guiffaut
TI  - Multi-label classification with deep learning techniques applied to the B-Scan images of GPR
JO  - Comptes Rendus. Physique
PY  - 2024
PB  - Académie des sciences, Paris
N1  - Online first
DO  - 10.5802/crphys.193
LA  - en
ID  - CRPHYS_2024__25_S1_A12_0
ER  - 
%0 Journal Article
%A Soukayna El Karakhi
%A Alain Reineix
%A Christophe Guiffaut
%T Multi-label classification with deep learning techniques applied to the B-Scan images of GPR
%J Comptes Rendus. Physique
%D 2024
%I Académie des sciences, Paris
%Z Online first
%R 10.5802/crphys.193
%G en
%F CRPHYS_2024__25_S1_A12_0
Soukayna El Karakhi; Alain Reineix; Christophe Guiffaut. Multi-label classification with deep learning techniques applied to the B-Scan images of GPR. Comptes Rendus. Physique, Online first (2024), pp. 1-16. doi : 10.5802/crphys.193.

[1] H. M. Jol Ground penetrating radar theory and applications, Elsevier, 2008

[2] T. Noreen; U. S. Khan Using pattern recognition with HOG to automatically detect reflection hyperbolas in ground penetrating radar data, 2017 International Conference on Electrical and Computing Technologies and Applications (ICECTA), IEEE (2017), 104620, pp. 1-6

[3] E. Temlioğlu; M. Dağ; R. Gürcan Comparison of feature extraction methods for landmine detection using ground penetrating radar, 2016 24th Signal Processing and Communication Application Conference (SIU), IEEE (2016), pp. 665-668 | DOI

[4] W. A. Wahab; J. Jaafar; I. Mohd Yassin; M. R. Ibrahim Interpretation of Ground Penetrating Radar (GPR) image for detecting and estimating buried pipes and cables, 2013 IEEE International Conference on Control System, Computing and Engineering, IEEE (2013), pp. 361-364 | DOI

[5] B. Walker; L. Ray Multi-class crevasse detection using ground penetrating radar and feature-based machine learning, IGARSS 2019-2019 IEEE International Geoscience and Remote Sensing Symposium, IEEE (2019), pp. 3578-3581 | DOI

[6] I. Giannakis; A. Giannopoulos; C. Warren A machine learning scheme for estimating the diameter of reinforcing bars using ground penetrating radar, IEEE Geosci. Rem. Sens. Lett., Volume 18 (2020) no. 3, pp. 461-465 | DOI

[7] N. Barkataki; S. Mazumdar; P. B. Devi Singha; J. Kumari; B. Tiru; U. Sarma Classification of soil types from GPR B scans using deep learning techniques, 2021 International Conference on Recent Trends on Electronics, Information, Communication & Technology (RTEICT), IEEE (2021), pp. 840-844 | DOI

[8] M. Sezgin; M. N. Alpdemir Classification of Buried Objects Using Deep Learning on GPR Data, 2023 IEEE International Conference on Advanced Systems and Emergent Technologies (IC_ASET), IEEE (2023), pp. 01-05 | DOI

[9] C. Maas; J. Schmalzl Using pattern recognition to automatically localize reflection hyperbolas in data from ground penetrating radar, Comput. Geosci., Volume 58 (2013), pp. 116-125 | DOI

[10] XLIM Institute Time Electromagnetic Simulator - Finite Difference Time Domain (Software Developed in Limoges, France)

[11] A. Taflove; S. C Hagness; M. Piket-May Computational electromagnetics: the finite-difference time-domain method, The Electrical Engineering Handbook, Volume 3, Academic Press, Burlington, 2005, pp. 629-670 | DOI

[12] V. Ciarletti; A. Herique; J. Lasue et al. CONSERT constrains the internal structure of 67P at a few metres size scale, Mon. Not. Roy. Astron. Soc., Volume 469 (2017) no. Suppl_2, p. S805-S817 | DOI

[13] R. P. K. Reddy; C. Nagaraju; I. R. Reddy Canny scale edge detection, 2015 (https://www.researchgate.net/publication/319701466_Canny_Scale_Edge_Detection)

[14] N. Barkataki; B. Tiru; U. Sarma A CNN model for predicting size of buried objects from GPR B-Scans, J. Appl. Geophys., Volume 200 (2022), 104620 | DOI

[15] D. Sarwinda; R. H. Paradisa; A. Bustamam; P. Anggia Deep learning in image classification using residual network (ResNet) variants for detection of colorectal cancer, Procedia Comput. Sci., Volume 179 (2021), pp. 423-431 | DOI

[16] H. H. Tan; K. H. Lim Vanishing gradient mitigation with deep learning neural network optimization, 2019 7th international conference on smart computing & communications (ICSCC), IEEE (2019), 60, pp. 1-4 | DOI

[17] K. Simonyan; A. Zisserman Very deep convolutional networks for large-scale image recognition (2014) (preprint arXiv:1409.1556) | DOI

[18] Z. Zhang Improved adam optimizer for deep neural networks, 2018 IEEE/ACM 26th international symposium on quality of service (IWQoS), IEEE (2018), pp. 1-2 | DOI

[19] A. Tato; R. Nkambou Improving adam optimizer, 2018 (https://openreview.net/forum?id=HJfpZq1DM)

[20] Y.-J. Cao; L.-L. Jia; Y.-X. Chen et al. Recent advances of generative adversarial networks in computer vision, IEEE Access, Volume 7 (2018), pp. 14985-15006 | DOI

[21] N. Jmour; S. Zayen; A. Abdelkrim Convolutional neural networks for image classification, 2018 international conference on advanced systems and electric technologies (IC_ASET), IEEE (2018), pp. 397-402 | DOI

[22] J. Read; F. Perez-Cruz Deep learning for multi-label classification (2014) (preprint arXiv:1502.05988) | DOI

[23] D. A Van Dyk; X.-L. Meng The art of data augmentation, J. Comput. Graph. Stat., Volume 10 (2001) no. 1, pp. 1-50 | DOI

[24] C. Shorten; T. M. Khoshgoftaar A survey on image data augmentation for deep learning, J. Big Data, Volume 6 (2019) no. 1, 60 | DOI

[25] B. Hu; C. Lei; D. Wang; S. Zhang; Z. Chen A preliminary study on data augmentation of deep learning for image classification (2019) (preprint arXiv:1906.11887) | DOI

[26] Y. Bai; E. Yang; B. Han et al. Understanding and improving early stopping for learning with noisy labels, NIPS’21: Proceedings of the 35th International Conference on Neural Information Processing Systems (Advances in Neural Information Processing Systems), Volume 34, Curran Associates, Inc. (2021), pp. 24392-24403

[27] M. Mahsereci; L. Balles; C. Lassner; P. Hennig Early stopping without a validation set (2017) (preprint arXiv:1703.09580) | DOI

Cité par Sources :

Commentaires - Politique