Journal of Xidian University ›› 2018, Vol. 45 ›› Issue (6): 92-98.doi: 10.3969/j.issn.1001-2400.2018.06.016

Previous Articles     Next Articles

Spatial sensing matrix learning for PolSAR image classification

SUN Chen;CHENG Liye   

  1. (Science and Technology on Reliability Physics and Application of Electronic Component Lab., the Fifth Electronics Research Institute of Ministry of Industry and Information Technology, Guangzhou 510610,  China)
  • Received:2017-12-16 Online:2018-12-20 Published:2018-12-20

Abstract: To deal with the large scale problem, a classifier based on the spatial sensing matrix and Least Squares Support Vector Machine (LS-SVM) is proposed for the polarimetric synthetic aperture radar (PolSAR) image. Inspired by the compressive sensing theory, a spatial sensing matrix is designed, which is equal to the product of the measurement matrix and the kernel matrix. The discriminative sensing matrix is proposed to largely reduce the scale of the optimization problem. Then, by taking the special properties of the polarimetric data and spatial information into account, we propose a spatial-Wishart dictionary to reduce the noise of speckle. Finally, the compressive sensing inspired classifier is constructed and the sparse support vector coefficients are achieved. Classification accuracy and spatial consistency of the proposed classifier is superior to those of other classifiers.

Key words: compressive sensing, polarimetric synthetic aperture radar image classification, discriminative sensing matrix, spatial-Wishart kernel

CLC Number: 

  • TP751