Advanced EMI Models and Classification Algorithms: The Next Level of Sophistication to Improve Discrimination of Challenging Targets

Dr. Fridon Shubitidze | Dartmouth College

MR-2225

Objective

The objective of this project was to enhance the detection, localization, and classification of small and/or deeply buried unexploded ordnance (UXO) by developing fast, noise-tolerant Electromagnetic Induction (EMI) data preprocessing and inversion approaches and by extending the detection range and spatial resolution of next-generation EMI systems using different combinations of transmitter coils and adjusting the coils in both direction and current amplitude. In a real field, electromagnetic responses from actual targets are mixed with noise due to both the instrument and the environment. This project studied two ways to increase the depth at which buried UXO can be detected and characterized: 1) increasing the signal amplitude by using high-power and/or large transmitter loops, and 2) decreasing the noise level through the use of advanced models for EMI data preprocessing and classification.

Back to Top

Technical Approach

This project provided efficient methodologies and practical approaches for improving the detection and classification of small and deeply buried targets at live UXO sites. First, enhancing EMI data preprocessing and minimizing noise by extending and adapting advanced EMI models—including the Orthonormalized Volume Magnetic Source (ONVMS) model, joint diagonalization (JD), multiple signal classification (MUSIC), and a combined JD-MUSIC approach—to existing system hardware was considered to extend detection range and spatial resolution. Second, current EMI sensor hardware was modified and updated to enhance the capability to detect and single out EMI signals from buried targets. The combined software/hardware suite was used to extract classification features from small and/or deep targets, and its performance was evaluated using test-stand and blind data sets. 

Back to Top

Results

Data analyses using advanced EMI models were conducted for test-stand data sets collected with the MetalMapper and the 5 ´ 5 and 3D-2 ´ 2 versions of the TEMTADS sensor. An updated system (a high-power 2 ´ 2 TEMTADS system with 14 A transmitter currents) was deployed and tested at the blind and small-munitions sites at Aberdeen Proving Ground in Maryland. The data were processed using a combined ONVMS-DE approach that extracted intrinsic and extrinsic parameters from targets and ultimately classified each one as either UXO or non-UXO. Excellent classification results were obtained at both grids for all anomalies larger than 20 mm.

Back to Top

Benefits

The results clearly demonstrated that the combined hardware/software approaches developed by this project could detect, locate, and classify buried small and/or deep UXO. The advanced EMI models described here shall remain the preferred approach for analyzing small and/or deep buried targets using EMI data. Not only was the combined approach able to detect and discriminate all shallow targets, it also managed to detect and classify small targets buried as deep as 20 times the target diameter. The high-power transmitter approach developed here has been implemented in a state-of-the-art instrument, a commercially available version of the 2 ´ 2 TEMTADS called the Geometrics mini MetalMapper.

Back to Top

Points of Contact

Principal Investigator

Dr. Fridon Shubitidze

Dartmouth College

Phone: 603-646-3671

Program Manager

Munitions Response

SERDP and ESTCP

Share