A System Dynamics Approach to Buried Mine/Unexploded (UXO) Detection and Identification

Current landmine detection approaches use very little object specific information from the sensor signal. A drawback is high false alarm rates due to large overlaps of the limited feature sets used to distinguish mines from clutter. In this project, we will focus on a model-based approach to feature extraction and pattern recognition for the acoustic-seismic mine detection platform built by University of Mississippi. The acoustic-seismic mine detection approach has shown a lot of promise in detecting low-metallic (plastic) mines that are considered hard to detect by conventional metal detectors and GPR. The motivation of our approach is that by imposing mine/soil model under acoustic-seismic excitation we can perform more robust system identification. This, in turn, will help us extract physically meaningful features for statistical learning and target recognition under limited but noisy data. The main objective of the proposed effort is to investigate candidates of ISSD model that is most suitable for use of mine detection. The requirements for such models are not only accuracy of the models but also identifiability of the models given inquiry signal and sensor data. Our second objective is to develop a parameter estimation method, in parallel with the ISSD model selection. The method is to be used to determine from the collected inquiry signal and detector data any model coefficients that are not known during mine detection operations. The project team consists of Scientific Systems Company, Inc. (SSCI) as the prime contractor, and the University of Mississippi as the sub-contractor.