Many Department of Defense (DoD) sites are contaminated by munitions and explosives of concern that are difficult to clean up because commercially available technology is inadequate in forested areas and in rugged terrain. This project demonstrated the Man Portable Vector (MPV), a handheld technology designed to extend the classification performance of the latest vehicle- and cart-based geophysical platforms to sites where vegetation or terrain limit access to these platforms. The MPV technology was evaluated as part of two ESTCP live-site demonstrations. The demonstration objectives were to show that the technology could be effectively utilized to map metallic object contamination and to reliably characterize the buried items as either potentially hazardous munitions and explosives of concern, or non-hazardous shrapnel, range scrap, or cultural debris.

Technology Description

The MPV sensor is a handheld, wide-band, time-domain, electromagnetic induction (EMI) sensor. The sensor head comprises a 50-cm diameter transmitter loop and an array of five receivers that measure the three components of the secondary magnetic fields induced by the transmitter. The MPV is field-programmable through a user interface that also provides immediate data feedback for quality control and target localization. The technology was initially tested by the Engineer Research and Development Center (ERDC)-Cold Regions Research and Engineering Laboratory (CRREL) under SERDP project MR-1443. The second-generation MPV demonstrated here was specially designed to improve maneuverability and ruggedness and was tested under ESTCP project MR-201158. Both MPVs were fabricated by G&G Sciences.

Generally operated by a two-person team, the MPV can be utilized in dynamic survey mode to map and locate potential targets, and in cued interrogation mode to characterize detected items. In cued mode, the sensor is placed at multiple locations to provide different “looks” at the buried target and is kept stationary to maximize data quality. The multiple soundings collected in cued-mode are jointly interpreted through geophysical inversion, which requires accurate positioning for accurate recovery of intrinsic target parameters.

The accuracy of standard positioning methods, such as GPS and roving laser rangers, may be compromised by thick canopy or the presence of obstacles. The MPV overcomes this limitation with a dedicated, local positioning system based on a portable receiver station that monitors the primary transmitter field from the MPV–like a beacon. This feature returns relative location estimates with centimeter-level accuracy out to a range of 4 meters and thus provides accurate positioning in most environmental conditions, including dense forest and under thick forest canopies.

Demonstration Results

This project included two demonstrations. The first demonstration was conducted at the former Spencer Artillery Range (SR) near Spencer, Tennessee, in June 2012. The MPV and a number of other advanced sensor technologies were tested on 1.3 acres of a flat open field area, where approximately 300 targets were encountered and classified. Portable systems were also tested in a forest where 700 targets were reacquired for classification. The second demonstration took place at the former Camp George West (GW) on the steep sides of Green Mountain in Lakewood, Colorado, in October 2012. MPV data were collected in full-coverage dynamic survey mode on approximately 2 acres and in cued-interrogation mode over approximately 500 anomalies selected from the full coverage data. The primary performance metrics were the probabilities of detection and correct classification.

At SR, classification of MPV cued-interrogation data resulted in the correct identification of 99% of the buried targets of interest (TOI) with rejection of 85% of the clutter. In most cases, the caliber of the munitions was correctly inferred from the MPV data. One seeded item, which was buried in a “fragmentation pit,” was incorrectly labeled as non-hazardous. The dynamic area was surveyed with 95% coverage and detection of all TOI. Classification was applied in two stages. The first stage used dynamic data as a pre-screener and eliminated the need for cued-interrogation on 50% of the anomalies. In the second stage, cued-data were collected and used for classification of the remaining anomalies. Using this hybrid approach, all TOI were found while rejecting 90% of the clutter.

At the GW site, the detection survey achieved 99% coverage and localization of all seeded items. Two independent classification studies were performed: one based on the full-coverage (dynamic) data and the other on the cued-interrogation data. Analysis of the cued data resulted in correct classification and prediction of the target caliber for 100% of the target while rejecting 90% of the clutter. Due to favorable environmental conditions and the absence of any small caliber TOI, analysis of the dynamic data resulted in 98% correct classification and 85% clutter rejection (one target was missed in the dynamic data set due to mislabeling).

These results from SR and GW suggest that, under favorable conditions, dynamic data can be sufficient to classify some of the field anomalies without the need for cued interrogations, resulting in further reductions in data collection and analysis costs.

Implementation Issues

The MPV technology has now been successfully demonstrated at four sites in conditions including open field, mountainous terrain, and moderately dense forest. The demonstrations proved the concepts of detection and classification with a handheld technology at a prototype stage. The novelty of each study required the presence of an expert geophysicist to oversee operations and to provide immediate feedback on the data collection procedures and data quality. Although the prototype may appear to be fragile, surveys were completed without any issues. Troubleshooting would still require intimate knowledge of the sensor. The average daily production rates were 0.6-0.7 acres for detection survey and 125-135 targets in cued interrogation. Detection was most efficient when using a third operator to lay out survey lines.

Efforts toward technology transition included training of three new field operators that successfully collected high quality data. Classification has also been successfully performed by multiple analysts and on different software platforms. Technical documentation on data collection and analysis procedures has been published in demonstration reports (ESTCP MR-201005 and MR-201158). A user manual with standard operating procedures is being written.

New demonstrations are scheduled as part of the ongoing ESTCP live-site demonstration program under project MR-201228. The upcoming sites will test conditions specifically tailored for the MPV sensor, such as dense forests, and will employ commercial crews to accelerate technology transfer and obtain feedback from seasoned industry professionals.