We will investigate the use of multisensor data fusion in mine detection and classification. Several sensor-fusion architectures will be explored, each of which makes use of a neural network for identify fusion. A proprietary technique will be used to train the network. The Phase I effort will comprise a feasibility study of sensor fusion for mine detection culminating in a preliminary design for a prototype device. The preliminary design study will specify sensor suites, processing algorithms, estimates of the system's physical characteristics (size, weight, and required power), and an operational concept. In a subsequent Phase II effort this preliminary design will be refined, and a prototype system will be developed and evaluated.
