The efficient and reliable identification of unexploded ordnance (UXO) without the need to excavate large numbers of non-UXO is one of the Department of Defense’s most pressing environmental problems. Discriminating UXO from non-UXO items is more difficult when sensor data are contaminated with geological and routine acquisition noise. In regions of strong magnetic geology, for example, magnetic sensors detect anomalies that are geologic, as well as metallic, in origin. Successful detection of UXO in these environments requires detecting all dipole-like magnetic anomalies and identifying and discarding geologic anomalies, as well as targets resulting from high frequency acquisition errors during field activities. Beyond detection, the next level of difficulty lies in discriminating UXO from non-UXO items of similar genesis such as scrap metal.
The objective of this project was to develop a comprehensive approach for detection of UXO-like targets in the presence of geologic noise and discrimination between UXO and non-UXO through indirect shape information contained in the magnetic higher-order moments. The project was designed to address the most difficult magnetic conditions where (1) the data noise tends to mimic the amplitude and phase information of potential UXO, increasing the number of false positives and (2) significant background response from soils, geology, and urban structures entirely mask the sought targets from detection.
The first task was the continued development and testing of a new method for UXO anomaly detection using a Hilbert transform-based extended Euler deconvolution. During initial development of the algorithm through SERDP project MR-1414, positive results in magnetic environments demonstrated the feasibility of the technique when pre-processing is applied to the field data to manage the geologic response. Current developments included modification of the algorithm to include higher quality magnetic gradient data and development of a post-processing statistical technique for reducing the number of false picks associated with high-frequency noise and geologic interference.
In the second task, researchers focused on the difficulty of discriminating UXO from non-UXO items with real data when sensor data are strongly contaminated with geological and cultural noise. Successful detection of UXO in these magnetic environments requires detecting all dipole-like magnetic anomalies and identifying and discarding the geologic anomalies that drastically increase the number of false targets. This component of the project had a major emphasis on developing or modifying robust processing algorithms, not common to UXO remediation efforts, as necessary for separating strong magnetic interference and enhancing the response of desired targets. The final task involved taking UXO discrimination between UXO and non-UXO beyond the current dipole-based approaches by utilizing the higher order magnetic moments that encode shape information about buried targets.
Throughout the project, several robust inversion algorithms tailored to the complex nature of the solution space were developed, and these techniques were applied to both realistic synthetic scrap/UXO models, as well as high quality data from real targets.
Successful target detection was demonstrated in strongly magnetic environments using this approach. The results also confirmed that the thresholding of targets selected as potential UXO by extended Euler deconvolution must move beyond the limited information contained in the phase. Secondary processing of identified target anomalies has demonstrated that amplitude information is an essential component for successfully reducing many of the false alarms associated with strong geologic noise.
The robust processing algorithms for difficult magnetic data sets were able to separate the background response in strongly interfering environments, significantly enhancing the UXO/scrap response. Field examples included magnetic data collected in Helena Valley, the former Camp Sibert, the former Camp San Luis Obispo, and an underwater test site.
Recovery of higher-order magnetic moments containing target shape information is theoretically possible with appropriate inversion algorithms and high-quality data. However, these diagnostic parameters are likely unrecoverable from real data over UXO/scrap targets for practical discrimination based on the total-field magnetic response.
This project developed an efficient technique that incorporates both amplitude and phase information to detect magnetic dipoles (UXO and scrap) in difficult magnetic data. The Hilbert transform-based extended Euler deconvolution method detects these targets based on their structural index values. Additionally, a set of post-processing statistical techniques was developed to reduce the number of false picks from data noise and magnetic geology. The project also developed and applied a series of robust processing algorithms for enhancing UXO response. The algorithms are not common to UXO remediation efforts, and in this project, they have demonstrated great success at separating and enhancing the magnetic response of UXO and scrap from strongly interfering magnetic soils, geology, urban structures, and platform motion.