Target recognition in infra-red imagery using neural networks and machine learning

This paper describes work undertaken by British Aerospace (BAe) on the evaluation of neural network and machine learning classifier techniques for automatic recognition of land based targets in infra-red imagery. The input to the classifier was derived from a histogram segmentation process extracting regions of interest from infra-red (IR) imagery. A set of statistical features were calculated for each region to form a feature vector describing the region. These feature vectors were then used as the input to the classifier. Two neural classifiers were investigated, based upon the radial basis function and multi-layer perceptron networks, and two machine learning classifiers, based upon the ID3 and CN2 techniques. In order to assess the merits of these approaches, the classifiers were compared with a conventional classifier originally developed by British Aerospace (Systems and Equipment) Ltd, under contract to RARDE (Chertsey), for the purpose of infra-red target recognition. This conventional system was based upon a Schurman classifier which operated on data transformed using a Hotelling trace transform. The ability of the classifiers to perform practical recognition of real-world targets was evaluated by training and testing the classifiers on real imagery obtained from mock land battles and military vehicle trials