Abstract : Automatic classification of vehicles on infra-red images is a hard image processing problem. Because it is difficult to use a pattern recognition method based on models, an alternative approach, based on learning by example, is investigated. The approach involves neural network techniques. Experimental results show that classification rates around 92% can be obtained with a training base of 1000 examples.
Gilles Burel, Jean-Yves Catros. Neural Networks for Vehicle Detection and Classification. Battlefield Atmospherics Conference, Nov 1993, Las Cruces, United States. ⟨hal-03222635⟩