Learning Machine Identification of Ferromagnetic UXO Using Magnetometry

Abstract

The fundamental problem in applying geophysical mapping to locate unexploded ordnance (UXO) is distinguishing true UXO from non-UXO. Enhancing the accuracy of UXO detection has multiple benefits, especially in the areas of cost savings and safety. We investigated discrimination approaches using both magnetic field data and numerically modeled data. Libraries of total field magnetic (TFM) responses were calculated using finite element modeling for three UXO types found at a Montana National Guard training site. UXO model parameters were varied over ranges of azimuth, declination, and depth resulting in approximately 600 models per UXO type. The modeled responses of finite-element model (FEM) and actual TFM field data were then used as training data in discrimination and classification approaches comparing neural networks (NN), random forests (RF), and support vector machines (SVMs). The prediction targets in the training process comprised three classes: 1) binary [UXO or noninteresting object (NIO)]; 2) multiclass (UXO round type and NIO); and 3) classes derived from multiclass self-organizing feature map (SOFM) analysis. The multiclass SOFM targets generated from site-specific field data were found to be optimal for UXO discrimination. The best performing combination of class selection types using recentered data for UXO detection rates of 100% resulted in a false alarm rate (FAR) of 28%. Index Terms-Classification, finite-element model (FEM), magnetics, neural networks (NN), random forest (RF), support vector machine (SVM), unexploded ordnance (UXO).

Sensitivity Analysis and Classification Algorithms Comparison for Underground Target Detection

“…These parameters are related to the physical characteristics of the target, and the classification results can be used as a basis for determining whether the target is the target of interest. The frequently used data classification methods include support vector machine (SVM) [27],random forest [27], neural network [27], [28], k-neighbor [24] and unsupervised weighted-pair group method with averaging (WPGMA) algorithm [26].…”

“…The model parameters are corrected iteratively through the ground truth, so that all targets of interest can be identified correctly [29]. Bray and Link utilized FEM model response and actual TFM field data as training data in discrimination and classification approaches, and compared the classification performance of neural networks, random forests, and support vector machines [27]. In [25], the detailed steps of UXO classification procedure using the advanced EMI sensors and models are presented along with the processing and analysis approaches that are used to generate a prioritized dig list.…”

“…Support Vector Machine algorithm is developed from choosing the best hyperplane in linear classification. In this paper the training data in sample space is nonlinear separable [22], [27].…”

“…where E µ r and E R are the mathematical expectation of relative permeability µ r and the sphere's radius R, respectively. V µ r is the variance of µ r , and V R is the variance of R. Plugging the value obtained in (27) and (9) into equations (14) (15) (16) and (17), the output uncertainty of the model and the contribution of each variable to output uncertainty can be obtained:…”

Transient Electromagnetic Voltage Imaging of Dense UXO-Like Targets Based on Improved Mathematical Morphology

“…The main survey methods for UXO include magnetic survey [2][3][4], electromagnetic method [5], [6] and ground penetrating radar (GPR) [7]. The magnetic survey is widely used to find UXO when the survey devices are mounted on the ground, in the water or the air.…”