Conducting unexploded ordnance (UXO) surveys using Global Positioning System (GPS)-guided geophysical mapping techniques has rapidly become the industry standard. Despite the investment in creating improved data, the use of physics-based signature analysis algorithms and analyses invariably requires digging 5-100 items for ordnance recovered intact. Employing either magnetometry or electromagnetic induction (EMI) sensors also requires physically clearing UXO ranges of vegetation and large debris. In SERDP SEED project MR-1354, the project team investigated several machine learning approaches to automate the analysis process and develop classification approaches based on the physical predictions of the dipole fitting routines used in analyzing magnetometry data. These included Artificial Neural Networks (ANN) and Genetic Ensemble Feature Selection (GEFS) approaches trained on ground-truthed magnetometry survey data to analyze and classify blind data. In each case, the machine learning approaches gave inferior results to those of a human analyst interactively analyzing targets using dipole fitting routines and visual cues to make classification predictions. Researchers turned to resampling the raw data arrays, using a machine learning approach to select candidate targets, then analyzing those targets applying shape function information and shape filters resident in an existing commercial software product, Feature Analyst. Working with vehicular Multi-Sensor Towed Array Detection System (MTADS) survey data from a ground artillery range, this approach dramatically reduced the false alarm rate over that of the human analyst. This analysis, designed to reduce false positives, unfortunately also mildly increased the false negative rate.

The objective of this project was to develop a more sophisticated automated screening approach (i.e., target picker) that increases the inclusion of all viable UXO candidates, while continuing to reject non-UXO. Specific objectives included: (1) further evolution of the characterization of size and shape parameters for use by the pattern recognition algorithms, (2) development of a probabilistic-based ranking system that can be adjusted to either maximize clutter rejection or to reduce false negatives, and (3) investigation of the fusion of physics-based and size- and shape-based parameters.

Technical Approach

This project investigated applications of the sophisticated automated screening approach for UXO classification to other types of magnetometry data, including airborne and marine data sets. Researchers considered other types of mapping approaches (gradiometric and analytic signal) to generate images containing shape information for pattern recognition algorithms. Other sensor data sets included EMI data (single or multiple time gates, or frequency-domain data). Range data of varying complexity in terms of target density, ordnance types, and geologically difficult terrain were evaluated.


The two-pass workflow in Feature Analyst, with the Target Picker and Target Ranker modules operating separately, successfully separated the Regions of Interest (ROIs) into likely UXO and targets unlikely to be UXO. The target ranker routinely batch processed the entire airborne dataset and ranked all ROIs. Using a conservative threshold of 0.20 provided a UXO dig list with about 1,300 entries. This number of entries is similar to the dig list generated by the MTADS DAS analyst using physics-based algorithms and hands-on analysis of each individual target. The latter process required approximately 35 hours of analyst time. 19 of 19 known UXO were correctly identified by the automated target ranker. The analytical signal processing of the magnetometer data was important, particularly for the airborne data sets. Using the analytical signal data also has the advantage that it locates the coordinates correctly for the individual targets. When working with the dipole presentations, only the positive lobes of the dipole signatures are used for training because various orientations of the individual targets creates such a variable dipole image when the negative part of the signature is included that it is not practical to try to train on the full dipole signature.


The benefits of this approach include the automation of target analysis processes to the best extent possible and improvement in the ability to reject clutter targets found during UXO cleanup, while retaining the ability to capture intact ordnance. It is likely that the greatest successes from this project will be realized by combining the approaches developed in this effort with the use of physics-based algorithm classification in some type of cooperative analysis approach. It is unlikely that an ultimate best-analysis approach will be accomplished without the provision for a human analyst in the process (at least to evaluate quality assurance). Future work will focus on the continued development and refinement of Target Picker and Target Ranker using data from challenging sites, specifically that from Isleta, Kirtland Precision Bombing Range. Vehicular and airborne survey MTADS data will be utilized along with EM array survey data for a comprehensive assessment.

  • Image-based ,

  • Classifier ,

  • Magnetometer ,

  • Analysis ,

  • Machine Learning