Classification of UXO Using Convolutional Networks Trained on a Limited Dataset

Lebbad, Anderson; Nataraj, C.

doi:10.1109/icmla.2017.000-1

Cited by 5 publications

(5 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As reported in Ref. [14], classification of mortars from the set in Figure 2 was found to be 97% accurate. Current work is focused on expanding the image database, testing the methods on additional types of ordnance, and enabling the classification of partially-occluded ordnance and ordnance in field conditions.…”

Section: Robotic Manipulatorsupporting

confidence: 75%

“…A more recent project, which is co-led by Dr. C. "Nat" Nataraj at VU, focuses on the development of an automated ordnance classification system to aid EOD technicians [14]. In typical EOD missions, classification of the ordnance that is being dealt with is of critical importance as it dictates how the EOD technician will approach the situation.…”

Section: Automated Ordnance Classification Support Systemmentioning

confidence: 99%

“…The concept for this classification support system is that an EOD technician will take a picture of a piece of ordnance and the system will return potential ordnance matches using a ConvNet classifier [14]. While these techniques are well established in machine learning, the availability of ordnance image datasets is limited or non-existent.…”

Section: Automated Ordnance Classification Support Systemmentioning

confidence: 99%

See 2 more Smart Citations

Mechatronics for Humanitarian Explosive Ordnance Disposal in Cambodia

Clayton

2018

Mechanical Engineering

View full text Add to dashboard Cite

Over the past six years, researchers at Villanova University (VU) and the Golden West Humanitarian Foundation (GWHF) have developed an integrated research and educational program focused on the use of mechatronics and robotics in humanitarian explosive ordnance disposal (EOD) and mine action. In the following article, I will talk about this program, discuss two ongoing projects - a low-cost EOD robot and an automated ordnance identification system - and talk about how we have successfully integrated students in the work. There are many opportunities for the DSCD community to get involved in this area and hopefully this article will pique your interest.

show abstract

Section: Robotic Manipulatorsupporting

confidence: 75%

Section: Automated Ordnance Classification Support Systemmentioning

confidence: 99%

Section: Automated Ordnance Classification Support Systemmentioning

confidence: 99%

See 1 more Smart Citation

Mechatronics for Humanitarian Explosive Ordnance Disposal in Cambodia

Clayton

2018

Mechanical Engineering

View full text Add to dashboard Cite

show abstract

“…An experiment of larger scale was conducted using a Convolution Neural Network in order to classify images of mortars taken from different angles [18]. The dataset was made up by capturing 3672 images of 46 individual objects in lab conditions which belong to seven different types of mortars.…”

Section: Related Work 21 Uxo Recognition From Visual Contentmentioning

confidence: 99%

Automatic Visual Recognition of Unexploded Ordnances Using Supervised Deep Learning

Begkas

Giannakeris

Ioannidis

et al. 2022

Proceedings of the 2022 International Conference on Multimedia Retrieval

View full text Add to dashboard Cite

Unexploded Ordnance (UXO) classification is a challenging task which is currently tackled using electromagnetic induction devices that are expensive and may require physical presence in potentially hazardous environments. The limited availability of open UXO data has, until now, impeded the progress of image-based UXO classification, which may offer a safe alternative at a reduced cost. In addition, the existing sporadic efforts focus mainly on small scale experiments using only a subset of common UXO categories. Our work aims to stimulate research interest in image-based UXO classification, with the curation of a novel dataset that consists of over 10000 annotated images from eight major UXO categories. Through extensive experimentation with supervised deep learning we uncover key insights into the challenging aspects of this task. Finally, we set the baseline on our novel benchmark by training state-of-the-art Convolutional Neural Networks and a Vision Transformer that are able to discriminate between highly overlapping UXO categories with 84.33% accuracy. CCS CONCEPTS• Computing methodologies → Object recognition; Visual content-based indexing and retrieval.

show abstract

“…These systems are widely used to detect targets in flat and large areas, and they can achieve rapid positioning and characterization of underground targets. In classification, various algorithms like support vector machines, deep learning, and convolutional networks [21][22][23] have been used to distinguish UXOs from harmless targets. In an inversion, some iterative algorithms are usually used to locate and characterize targets, such as the Gauss-Newton [24,25] and differential evolution (DE) algorithms [26,27].…”

Section: Introductionmentioning

confidence: 99%

Fast Localization and Characterization of Underground Targets with a Towed Transient Electromagnetic Array System

Wang

Zhang

Chen

et al. 2022

Sensors

View full text Add to dashboard Cite

A fast inversion algorithm combined with the transient electromagnetic (TEM) detection system has important significance for improving the detection efficiency of unexploded ordnance. The traditional algorithms, such as differential evolution or Gauss–Newton algorithms, usually require tens to thousands of iterations to locate the underground target. A new algorithm with a magnetic gradient tensor and singular value decomposition (SVD) to estimate the target position and characterization quickly and accurately is proposed in this paper. Two modes of magnetic gradient tensor are constructed to accurately locate shallow and deep targets, respectively. The SVD algorithm is applied to the responses to estimate the electromagnetic characteristics of the target quickly and accurately. To verify the performance of the proposed algorithm, a towed TEM sensor is designed, which is constructed with three transmitting coils and nine three-component receiving coils arranged in a 3 × 3 array. Field experiments in survey and cued modes were taken to verify the performance of the proposed algorithm and the towed system. Results show that the magnetic gradient tensor algorithm proposed in this paper can accurately locate a single target within 2.0 m depth, and the error of depth is no more than 8 cm. Even for overlapping response of multi targets, the error of depth is no more than 12 cm. The underground target can be accurately characterized by the SVD algorithm. For targets with depths over 2.0 m, the signal-to-noise ratio of characteristic response estimated by SVD is higher than that of the traditional method. The proposed method needs approximately 40 ms, only 1% of the traditional one, considerably improving detection efficiency and laying a theoretical and experimental foundation for real-time data processing.

show abstract

Classification of UXO Using Convolutional Networks Trained on a Limited Dataset

Cited by 5 publications

References 5 publications

Mechatronics for Humanitarian Explosive Ordnance Disposal in Cambodia

Mechatronics for Humanitarian Explosive Ordnance Disposal in Cambodia

Automatic Visual Recognition of Unexploded Ordnances Using Supervised Deep Learning

Fast Localization and Characterization of Underground Targets with a Towed Transient Electromagnetic Array System

Contact Info

Product

Resources

About