Automatic classification of aerial imagery

The aerial imagery obtained from reconnaissance platform is voluminous and the defence forces rely on image information to perform intelligent tasks. The application of a well designed automatic image classifier would enhance the end results of different high level applications thereby abridging the effort of a human analyst. Automatic image classifiers could be designed using a training data set for supervised learning or using an unsupervised learning. In this paper, a method, which combines both unsupervised and supervised methods of learning is proposed.