High-Order Neural Network-Based Ship Detection Algorithms Applied to SAR Imagery

Konferenz: EUSAR 2012 - 9th European Conference on Synthetic Aperture Radar
23.04.2012 - 26.04.2012 in Nuremberg, Germany

Tagungsband: EUSAR 2012

Seiten: 4Sprache: EnglischTyp: PDF

Persönliche VDE-Mitglieder erhalten auf diesen Artikel 10% Rabatt

de Nicolás-Presa, Jaime Martín; de la Mata-Moya, David; Jarabo-Amores, Maria-Pilar; Bárcena-Humanes, Jose-Luis; Palma-Vazquez, Angel (Signal Theory and Communications Department, Higher Polytechnic School, University of Alcala, 28805 Alcalá de Henares, Madrid, Spain)

Ship detection is traditionally carried out with patrol ships or aircrafts, with a limited coverage area and also limited by weather conditions. Synthetic aperture radars can surpass these limitations. In this paper, two neural network-based techniques are proposed and compared for detecting ships over a TerraSAR-X image. The first technique is based on a second-order neural network whereas the second one is based on the combination of a Zernike moments-based feature extractor and a Multilayer Perceptron. Good results are obtained with both techniques but the Zernike moments-based one gives rise to the best detection results with less processing time.