Classification of two ship targets using radar backscatter

This paper describes the implementation of a classifier which discriminates between ship targets, based on the radar backscatter from these targets. Three different classification methods were tested, namely correlation filters, peak extraction with a feed-forward neural network classifier and a feed-forward neural network which used the raw radar data to classify the ships. The correlation based method performed relatively poorly when compared to the two feature based methods. A maximum performance of 95.71% was obtained when using a feed-forward neural network (multilayer perceptron) with the raw segmented one-dimensional profile as input. The usage of peak detection as feature extraction and an MLP as classifier resulted in similar, but slightly poorer performance (95%).