A collective computation approach to automatic target recognition

A neural network associative memory is used to guide the template-matching process in a digital simulation model of a third-generation fuse. The templates are produced by analytic models of target interaction with several simultaneous sensors and are stored by means of the Hinton-Sejnowski formula in a symmetrically cross-coupled (Hopfield) network. A description is given of some salient features of network design and simulation. The performance gain that can be achieved with this collective computation approach is limited mainly by the size of the network.<
>