A Modified AdaBoost Algorithm with New Discrimination Features for High-Resolution SAR Targets Recognition

Kun CHEN  Yuehua LI  Xingjian XU  Yuanjiang LI  

IEICE TRANSACTIONS on Information and Systems   Vol.E98-D   No.10   pp.1871-1874
Publication Date: 2015/10/01
Publicized: 2015/07/21
Online ISSN: 1745-1361
DOI: 10.1587/transinf.2015EDL8090
Type of Manuscript: LETTER
Category: Pattern Recognition
synthetic aperture radar (SAR),  automatic target recognition (ATR),  adaptive boosting,  high-resolution,  

Full Text: PDF>>
Buy this Article

In this paper, we first propose ten new discrimination features of SAR images in the moving and stationary target acquisition and recognition (MSTAR) database. The Ada_MCBoost algorithm is then proposed to classify multiclass SAR targets. In the new algorithm, we introduce a novel large-margin loss function to design a multiclass classifier directly instead of decomposing the multiclass problem into a set of binary ones through the error-correcting output codes (ECOC) method. Finally, experiments show that the new features are helpful for SAR targets discrimination; the new algorithm had better recognition performance than three other contrast methods.