PolSAR Image Segmentation Based on the Modified Non-negative Matrix Factorization and Support Vector Machine

To improve polarimetric synthetic aperture radar (PolSAR) imagery segmentation accuracy, a modified non-negative matrix factorization algorithm based on the support vector machine is proposed. Focusing on PolSAR remote sensing images, the modified non-negative matrix factorization algorithm with the neurodynamic optimization achieves the image feature extraction. Compared with basic features, such as the basic backscatter coefficient, structuring more targeted localization non-negative character fits better for the physical significance of remote sensing images. Furthermore, based on the new constructive features, a support vector machine is employed for remote sensing image classification, which remedies the small sample training problem. Simulation results on PolSAR image classification substantiate the effectiveness of the proposed approach.

[1]  Moncef Gabbouj,et al.  Evolutionary RBF classifier for polarimetric SAR images , 2012, Expert Syst. Appl..

[2]  Turker Ince,et al.  Unsupervised classification of polarimetric SAR image with dynamic clustering: An image processing approach , 2010, Adv. Eng. Softw..

[3]  Wen Hong,et al.  An Unsupervised Segmentation With an Adaptive Number of Clusters Using the $SPAN/H/\alpha/A$ Space and the Complex Wishart Clustering for Fully Polarimetric SAR Data Analysis , 2007, IEEE Transactions on Geoscience and Remote Sensing.

[4]  Hai Jin,et al.  Color Image Segmentation Based on Mean Shift and Normalized Cuts , 2007, IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics).

[5]  David A. Clausi,et al.  Unsupervised Polarimetric SAR Image Segmentation and Classification Using Region Growing With Edge Penalty , 2012, IEEE Transactions on Geoscience and Remote Sensing.

[6]  Kei Suwa,et al.  Analysis on the Resolution of Polarimetric Radar and Performance Evaluation of the Polarimetric Bandwidth Extrapolation Method , 2013, IEEE Transactions on Geoscience and Remote Sensing.

[7]  Jun Wang,et al.  Single point iterative weighted fuzzy C-means clustering algorithm for remote sensing image segmentation , 2009, Pattern Recognit..

[8]  Chih-Jen Lin,et al.  A comparison of methods for multiclass support vector machines , 2002, IEEE Trans. Neural Networks.

[9]  Yang Shao,et al.  Comparison of support vector machine, neural network, and CART algorithms for the land-cover classification using limited training data points , 2012 .

[10]  Gabriele Moser,et al.  A Textural–Contextual Model for Unsupervised Segmentation of Multipolarization Synthetic Aperture Radar Images , 2013, IEEE Transactions on Geoscience and Remote Sensing.

[11]  H. Sebastian Seung,et al.  Learning the parts of objects by non-negative matrix factorization , 1999, Nature.