DETECTION OF MINES AND MINELIKE TARGETS USING PRINCIPAL COMPONENT ANDNEURAL-NETWORK METHODS

This paper introduces a new system for real-time detection and classif ication of arbitrarily scattered surface-laid mines from multispectral imagery data of a minefield. The system consists of six channels whic h use various neural-network structures for feature extraction, detect ion, and classification of targets in six different optical bands rang ing from near UV to near IR, A single-layer autoassociative network tr ained using the recursive least square (RLS) learning rule was employe d in each channel to perform feature extraction. Based upon the extrac ted features, two different neural-network architectures were used and their performance was compared against the standard maximum likelihoo d (ML) classification scheme. The outputs of the detector/classifier n etwork in all the channels were fused together in a final decision-mak ing system. Two different final decision making schemes using the majo rity voting and weighted combination based on consensual theory were c onsidered. Simulations were performed on real data for six bands and o n several images in order to account for the variations in size, shape , and contrast of the targets and also the signal-to-clutter ratio, Th e overall results showed the promise of the proposed system for detect ion and classification of mines and minelike tagets.