Research on Modified SVM for classification of SAR images

DOI

10.2991/fmsmt-17.2017.234How to use a DOI?

Keywords

Synthetic radar images (SAR), SVM, Contourlet transform.

Abstract

Classification of synthetic radar images (SAR) is an emerging area especially with the advent of state of the art satellite image techniques. An SVM based texture analysis and classification utilizing the PCA for dimensionality reduction of SAR images has been presented in this paper to categorize the given SAR image into the water and urban areas. The experimentation has been conducted on 40 SAR images and the feature set size is 15. Finally, most effective 5 texture features are shortlisted for the classification of SAR images and accuracy is calculated by Specificity and Sensitivity test. The results obtained from test images give an accuracy of 94% for image classification. To make the algorithm adaptable, these textural features are reduced using principal component analysis (PCA), and principal components are used for classification purposes powered by a support vector machine classifier. The well known multiresolution approximation technique contourlet transform has been utilized in this paper to pre-process the input image in the frequency domain effectively and also to select the most significant features in the frequency domain. The proposed technique has been compared with conventional techniques such as the PCA and SVM in stand-alone approaches and the classification accuracy is increased along with the drastic reductions in the computation time.

Copyright

© 2017, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Peng Zhou
AU  - Gang Guo
AU  - Fu Xiong
PY  - 2017/04
DA  - 2017/04
TI  - Research on Modified SVM for classification of SAR images
BT  - Proceedings of the 2017 5th International Conference on Frontiers of Manufacturing Science and Measuring Technology (FMSMT 2017)
PB  - Atlantis Press
SP  - 1193
EP  - 1199
SN  - 2352-5401
UR  - https://doi.org/10.2991/fmsmt-17.2017.234
DO  - 10.2991/fmsmt-17.2017.234
ID  - Zhou2017/04
ER  -