.png){kind=icon}
.png){kind=icon}
Return
A Modified Homotopy-Based Tensor Eigenpairs Algorithm for Remote Sensing Feature Extraction
.png){kind=icon}
PRE
AI
DOI：10.1109/LGRS.2021.3089153.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Independent component analysis (ICA) is one of the widely used techniques in remote sensing feature extraction. When selecting high-order statistics (HOS) as non-Gaussian metric, the determination of independent components (ICs) can be attributed to calculating the eigenpairs of HOS tensors of the mixed data. However, previous algorithms can only obtain approximate solutions for eigenpairs, and the accuracy of ICs may be unavoidably affected. Recently, a homotopy-based tensor eigenpairs (HTE) method that can obtain accurate solutions has been proposed. In this letter, we introduce it into the ICA field and further propose a modified version, termed modified HTE (MHTE). MHTE incorporates the concept of the projected Hessian matrix into HTE, which can further improve the accuracy of ICs and also intellectually determine the number of ICs. Experiments with both simulated image and remote sensing image demonstrate that it is more accurate and parameter adaptive than other compared feature extraction algorithms.
Keywords:
Tensors
Integrated circuits
Feature extraction
Remote sensing
Measurement
Optimization
Symmetric matrices
Eigenpairs
feature extraction
independent component analysis (ICA)
projected Hessian
symmetric tensors
Journal
IF:
16.4
Papers:
1.0W
Citations:
5.1K