Convolutional Autoencoder for Landmine Detection on GPR Scans

Picetti, Francesco; Testa, Giuseppe; Lombardi, Federico; Bestagini, Paolo; Lualdi, Maurizio; Tubaro, Stefano

doi:10.1109/tsp.2018.8441206

Cited by 20 publications

(16 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The system is trained on a large dataset of synthetic along with a few background-only real data and it reaches good performance in terms of detection accuracy. Finally, an alternative solution based on a one-class approach is proposed in [11].…”

Section: A Machine Learning For Geophysicsmentioning

confidence: 99%

“…In [9] the authors compare a neural network against logistic regression algorithms in discriminating between potential targets and clutter. Additionally, in [10], [11], the authors propose different CNNbased strategies for landmine detection in B-scans.…”

Section: Introductionmentioning

confidence: 99%

“…In this work we exploit a particular CNN architecture known as autoencoder to analyze GPR volumetric data acquired with multiple polarizations, building upon our recent findings reported in [11]. The proposed method is able to detect whether an investigated volume contains buried objects that are not coherent with the kind of soil being analyzed.…”

Section: Introductionmentioning

confidence: 99%

“…• This is the first solution jointly exploiting horizontal and vertical polarizations for landmine detection using CNNs, to the best of our knowledge. • The proposed solution works considering 3D volumetric data, rather than simple 2D B-scans as done in [10], [11]. • The proposed architecture is able to work in a crossdataset scenario, i.e., we can train the CNN on a dataset, and use it on another one with limited accuracy loss, thus making the solution robust to different environments.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Landmine Detection Using Autoencoders on Multi-polarization GPR Volumetric Data

Bestagini¹,

Lombardi²,

Lualdi³

et al. 2018

Preprint

Self Cite

View full text Add to dashboard Cite

Buried landmines and unexploded remnants of war are a constant threat for the population of many countries that have been hit by wars in the past years. The huge amount of human lives lost due to this phenomenon has been a strong motivation for the research community toward the development of safe and robust techniques designed for landmine clearance. Nonetheless, being able to detect and localize buried landmines with high precision in an automatic fashion is still considered a challenging task due to the many different boundary conditions that characterize this problem (e.g., several kinds of objects to detect, different soils and meteorological conditions, etc.). In this paper, we propose a novel technique for buried object detection tailored to unexploded landmine discovery. The proposed solution exploits a specific kind of convolutional neural network (CNN) known as autoencoder to analyze volumetric data acquired with ground penetrating radar (GPR) using different polarizations. This method works in an anomaly detection framework, indeed we only train the autoencoder on GPR data acquired on landmine-free areas. The system then recognizes landmines as objects that are dissimilar to the soil used during the training step. Experiments conducted on real data show that the proposed technique requires little training and no ad-hoc data preprocessing to achieve accuracy higher than 93% on challenging datasets.

show abstract

Section: A Machine Learning For Geophysicsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Landmine Detection Using Autoencoders on Multi-polarization GPR Volumetric Data

Bestagini¹,

Lombardi²,

Lualdi³

et al. 2018

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…The data are then processed through a learning technique, known as Convolutional Neural Networks (CNNs), to detect the presence of buried objects not coherent with the surrounding soil texture [36][37][38]. Compared to other supervised learning alternatives, CNNs, a deep learning technique particularly suited for computer vision and image classification tasks, determine the classification features in an automated way directly from the raw data, i.e., without the need for extensive manual data labelling.…”

Section: Introductionmentioning

confidence: 99%

Ballistic Ground Penetrating Radar Equipment for Blast-Exposed Security Applications

et al. 2020

Self Cite

View full text Add to dashboard Cite

Among all the forensic applications in which it has become an important exploration tool, ground penetrating radar (GPR) methodology is being increasingly adopted for buried landmine localisation, a framework in which it is expected to improve the operations efficiency, given the high resolution imaging capability and the possibility of detecting both metallic and non-metallic landmines. In this context, this study presents landmine detection equipment based on multi-polarisation: a ground coupled GPR platform, which ensures suitable penetration/resolution performance without affecting the safety of surveys, thanks to the inclusion of a flexible ballistic shielding for supporting eventual blasts. The experimental results have shown that not only can the blanket absorb blast-induced flying fragments impacts, but that it also allows for the acquisition of data with the accuracy required to generate a correct 3D reconstruction of the subsurface. The produced GPR volume is then processed through an automated learning scheme based on a Convolutional Neural Network (CNN) capable of detecting buried objects with a high degree of accuracy.

show abstract