U-HUNTER (also called “UXOHUNTER”) is a computer program that affords knowledge-based real-time sensor- fusion and display capabilities for detecting buried objects and materials of interest. U-HUNTER is intended especially for inferring the presence of buried unexploded ordnance and explosive waste from the readings of magnetic and electromagnetic sensors like those commonly used in geophysical surveys. U-HUNTER is also potentially adaptable to such other uses as detection of mines, medical imaging and diagnosis, detecting and monitoring buried pipes and cables, environmental monitoring, and geological surveys.

U-HUNTER can utilize data in a wide variety of formats. At present, it can accept and process the data from such sensors as magnetometers, syntheticaperture radar systems, and time-domain pulsed eddy-current sensors. UHUNTER has the potential for reducing (in comparison with prior sensor-data-analysis software) the rate of errors in the detection of anomalies (that is, unexploded ordnance or other phenomena of interest). U-HUNTER also has the potential for enabling better characterization and classification of detected anomalies. Envisioned applications of U-HUNTER include not only analysis of fresh data from continuing surveys but also re-evaluation of historical survey data to increase the level of certainty of detection of unexploded ordnance at sites where such surveys have already been performed. Fusion of multiple-sensor and/or multiple- survey data from the same site is also envisioned.

U-HUNTER contains unique subsystems that perform different functions, as follows:

  • A data-enhancement subsystem preprocesses input data to condition them for analysis. One of the most powerful features of U-HUNTER is a fully integrated programming interface within the data-enhancement subsystem. This interface makes it possible to develop custom algorithms to preprocess and display the data to be fed as input to the analysis-engine subsystem mentioned below. U-HUNTER is flexible enough to allow the import of any image or raw matrix data for analysis. This capability alone has proven to be most impressive in enabling users familiar with data from sensors of one type (e.g., magnetometers) to apply the analytical capabilities of U-HUNTER to data from sensors of another type (e.g., synthetic-aperture radar).

• An analysis-engine subsystem includes a version of ARTMAP, which is part of a previously developed neural-network software system. The analysis-engine subsystem is also expected to include ANFIS — a yet-to-be-developed program for implementing an adaptive neuro-fuzzy information software system for a secondary method of classification of anomalies.

  • A subsystem called “3DKB” is dedicated to facilitation of the assimilation of the items of a three-dimensional knowledge base. The 3DKB items comprise the knowledge built up by a user from ground-truth data for the purposes of training neural and neuro-fuzzy software components within U-HUNTER. Dialog-type displays enable users to load 3DKB items for training of the ARTMAP-type neural networks for classification of anomalies.

• A visualization subsystem generates displays of geophysical data on three-dimensional-appearing surface models and two-dimensional color maps. The maps provide a capability for rapid visual inspection of a surveyed area. Deep-blue dots are placed at locations of local maxima of and magenta dots are placed at locations of local minima of quantities of interest. The user can customize maximum and minimum-finding algorithms to obtain desired levels of discrimination. A subprogram called “3D Anomaly Viewer” enables users to see the shapes of found anomalies.

U-HUNTER has been found to perform robustly, both (1) when applied in real time at the survey-grid-interval level (2). U-HUNTER enables detection of anomalies at the 100-percent level, while keeping the incidence of false alarms down to the 2-percent level. It also provides for robust generalization to new sites without additional training or setup, and it minimizes the effect of reacquisition of data on postprocessing. It can perform robustly despite poor ground control; that is, it can tolerate errors of the order of grid intervals. Furthermore, U-HUNTER enables rapid synthetic resampling and interpolation of data to a spatial resolution relevant to the spatial distribution of unexploded ordnance, and enables resampling of data from multiple sensors to a common spatial scale.

Planned future development of UHUNTER includes an extension of datadisplay capabilities to provide for overlays of planimetric data, aerial photography, and geophysical data. Other planned enhancements include extension of analysis subsystems to accommodate additional alterative classification methods and development of an improved user interface to make the UHUNTER more accessible to the general geophysical community.

This work was done by Sandeep Gulati and Vijay Daggumati of Caltech for NASA’s Jet Propulsion Laboratory. For further information, access the Technical Support Package (TSP) free on-line at www.nasatech.com/tsp  under the Physical Sciences category.

This software is available for commercial licensing. Please contact Don Hart of the California Institute of Technology at (818) 393- 3425. Refer to NPO-20808.