Projection Pursuit Feature Analysis for Pan-Sharpened Multispectral Ikonos Imagery

Orbiting multispectral (MS) sensors can facilitate feature discrimination in clutter since natural clutter and man-made objects often differ in the energy they radiate across the electromagnetic spectrum. The projection pursuit technique (PP) has been previously proposed for assessing information content in large multivariate data sets such as hyperspectral (HS) imagery. Although the number of spectral bands is limited in MS data sets, this study investigates the suitability of PP for target detection. PP can highlight different features of interest in an image, improving and simplifying subsequent detection. This study uses two data sets: (1) 4 m MS IKONOS data and (2) pan-sharpened 1 m IKONOS MS imagery created by fusing the 4m MS and the associated 1 m panchromatic image sets. It is shown that PP based on the information divergence index can facilitate detection of certain targets, and emphasize features in MS data. This paves the way for an automated target detection and recognition system based around a PP preprocessing procedure