Article: Automatic target recognition in SAR images using quaternion wavelet transform and principal component analysis Journal: International Journal of Computational Vision and Robotics (IJCVR) 2017 Vol.7 No.3 pp.314 - 334 Abstract: Automatic target recognition (ATR) is the task of classifying sensed imagery from synthetic aperture radar (SAR) automatically into a canonical set of target classes. Here, a method to recognise different classes of military vehicles based on the combination of quaternionic wavelet transform (QWT) and principal component analysis (PCA) features is presented. To identify the certain region of SAR images, patches are extracted over the interest points detected from the SAR images. Then QWT features and PCA features are computed and combined for every patch. These extracted features are trained and classified using SVM. The performance of QWT is compared with two more multiresolution transforms such as ridgelet transform and log Gabor transform as well as the Scale and rotation-invariant interest point detector and descriptor, named speeded up robust features (SURF). Observations revealed that QWT outperforms the ridgelet transform, log-Gabor and SURF. The experimental evaluation is done using the MSTAR database. Inderscience Publishers - linking academia, business and industry through research

Title: Automatic target recognition in SAR images using quaternion wavelet transform and principal component analysis

Authors: S. Arivazhagan; R. Ahila Priyadharshini; L. Sangeetha

Addresses: Department of Electronics and Communication Engineering, Mepco Schlenk Engineering College, Sivakasi – 626 005, Tamil Nadu, India ' Department of Electronics and Communication Engineering, Mepco Schlenk Engineering College, Sivakasi – 626 005, Tamil Nadu, India ' Department of Electronics and Communication Engineering, Mepco Schlenk Engineering College, Sivakasi – 626 005, Tamil Nadu, India

Abstract: Automatic target recognition (ATR) is the task of classifying sensed imagery from synthetic aperture radar (SAR) automatically into a canonical set of target classes. Here, a method to recognise different classes of military vehicles based on the combination of quaternionic wavelet transform (QWT) and principal component analysis (PCA) features is presented. To identify the certain region of SAR images, patches are extracted over the interest points detected from the SAR images. Then QWT features and PCA features are computed and combined for every patch. These extracted features are trained and classified using SVM. The performance of QWT is compared with two more multiresolution transforms such as ridgelet transform and log Gabor transform as well as the Scale and rotation-invariant interest point detector and descriptor, named speeded up robust features (SURF). Observations revealed that QWT outperforms the ridgelet transform, log-Gabor and SURF. The experimental evaluation is done using the MSTAR database.

Keywords: SAR target recognition; interest points; patch extraction; quaternion wavelet transform; QWT; automatic target recognition; synthetic aperture radar; SAR images; principal component analysis; PCA; military vehicles; vehicle recognition; feature extraction; SVM; support vector machines.

DOI: 10.1504/IJCVR.2017.083449

International Journal of Computational Vision and Robotics, 2017 Vol.7 No.3, pp.314 - 334

Received: 16 Feb 2015
Accepted: 04 Jun 2015
Published online: 31 Mar 2017 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article

Title: Automatic target recognition in SAR images using quaternion wavelet transform and principal component analysis

Keep up-to-date