Neuromorphic Methods for Recognition of Compact Image Objects
Axel J. Pinz, Marek B. Zaremba, Horst Bischof, François A. Gougeon, Michel Locas · 1993
The issue of the recognition of tree species from high resolution aerial images is addressed in this paper. An approach based on the use of neural networks is presented and discussed in more detail. The networks perform classification and recognition operations on compact image objects, obtained by applying different tree isolation procedures. The recognition capabilities of two classes of networks, multilayer feedforward networks and holographic networks, are compared and some results of the research carried out in Austria and Canada, using aerial photographs and multispectral scanner images, are given.