Information-based sensor management for landmine detection using electromagnetic induction, ground-penetrating radar, and seismic sensors

An information-based sensor management framework is discussed that enables the automated tasking of a suite of sensors when detecting static targets. The sensor manager chooses the sensors to use and the grid-based locations to observe in order to maximize the expected information gain that will be obtained with each new sensor observation. Initially, sensor probabilities of detection and false alarm, Pd and Pf, are assumed to be known by the sensor manager. In a field setting, however, Pd and Pf cannot be known exactly, and so uncertainty modeling for Pd and Pf is also discussed. The sensor manager is tested on real landmine data using electromagnetic induction (EMI), ground-penetrating radar (GPR), and seismic sensors. A matched subspace detector is used to process the EMI data, an adaptive pre-screening algorithm based on the least mean squares (LMS) adaptive filter is used to process the GPR data, and whitening followed by an energy detector is used to process the seismic data. The sensor manager is able to detect the landmines more quickly and more effectively than an unmanaged, blind-search approach. Using all three sensor modalities also results in superior detection performance to that achieved by only a single sensing modality.