|
|
An Improved Adaptive Non-negative Eigenvalue Decomposition for Polarimetric Synthetic Aperture Radar |
Liu Gao-feng Li Ming Wang Ya-jun Zhang Peng Wu Yan |
National Key Laboratory of Radar Signal Processing, Xidian University, Xi’an 710071, China |
|
|
Abstract There are two problems for the Adaptive Non-Negative Eigenvalue Decomposition (ANNED). One is that the search solution to the Non-Negative Eigenvalue Decomposition (NNED) takes much time to obtain the volume scattering power by repeatedly calculating eigenvalues. The other is that the remainder covariance matrix of the ANNED may have negative eigenvalues which indicates that the decomposition result is invalid. For these two problems, an improved ANNED is proposed. Firstly, a fast solution to the NNED is derived by calculating the principal minor zeros of the remainder covariance matrix. Since it does not need to repeatedly calculate eigenvalues, the fast solution is faster than the search solution for calculating the volume scattering power. Secondly, the fast solution is also used to adjust the scattering powers of the ANNED to deal with the problem of the negative eigenvalue. The real-POLSAR-data experiment shows that the improved ANNED can markedly enhance the double-bounce scattering power and weaken the volume scattering power for urban areas, and enhance the accuracy of classification.
|
Received: 06 September 2012
|
|
Corresponding Authors:
Liu Gao-feng
E-mail: gaofengliu@mail.xidian.edu.cn
|
|
|
|
|
|
|