Decrease of sidelobe energy by shifts in original training images for synthetic aperture radar image recognition

Abstract

For synthetic aperture radar (SAR) image recognition, correlation filters were designed in our previous research by the nonzero‐ order joint‐transform correlator. Training images were placed at the presumed center in order to synthesize the reference function. However, this could not guarantee that the average cross correlation energy would be minimum. To improve this, training images are shifted pixel by pixel from left to right and top to bottom in this study. We synthesize the reference functions, hoping to find lower cross‐correlation energy. Finally, we obtain sharper correlation peaks and less sidelobe energy, which can increase the recognition ability. © 2004 Wiley Periodicals, Inc. Microwave Opt Technol Lett 41: 241–244, 2004; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/mop.20105