计算机科学 ›› 2015, Vol. 42 ›› Issue (Z6): 215-217.

• 模式识别与图像处理 • 上一篇    下一篇

SAR图像局部非线性分布特征及提取算法

管涛,于浩杰   

  1. 郑州航空工业管理学院计算机科学与应用系 郑州450015,郑州航空工业管理学院计算机科学与应用系 郑州450015
  • 出版日期:2018-11-14 发布日期:2018-11-14
  • 基金资助:
    本文受国家自然科学基金项目(41171341),河南省基础与前沿技术研究计划项目(132300410186),河南省教育厅科学技术研究重点项目资助

Local Nonlinear Distribution Feature and Extraction Algorithm for SAR Images

GUAN Tao and YU Hao-jie   

  • Online:2018-11-14 Published:2018-11-14

摘要: 谱聚类是当今机器学习领域的研究热点,大多数算法用于图像分割。由于谱聚类能够刻画数据在低维空间内的主要特性,因此分析了谱聚类表示特征的原理,构造了一种面向图像子块的非线性局部特征,提出了相应的特征提取算法,用于刻画SAR图像分块的性质。这些特征由谱聚类产生的若干特征值构成的向量组成,然后经过傅里叶变换得到,因而具有平移不变性。在计算的过程中,可以采用Nystrm等方法解决谱聚类中矩阵不可逆问题。为了避免减弱局部特性差异,在子块相似性计算中采用了明氏距离。实验验证了所提特征的有效性。

Abstract: Spectral clustering is a current research focus and most algorithms are applied to image segmentation.With the capacity of finding low dimension space of spectral clustering,this paper analyzed the principle of feature representation of spectral clustering,proposed a new local nonlinear distribution feature extracted from sub-blocks of SAR images and used to describe the properties of sub-blocks.These feature vectors are rotationally invariant and they are obtained via several steps.First,the initial feature vectors are obtained via spectral clustering and then they are transformed by discrete Fourier transform.We used Nystrm approach to compute the eigenvalues of spectral clustering.In order to avoid weakening the difference of local characteristics in sub-graphs,we adopted Minkowski distance to compute simi-larity among sub-graphs.The efficiency of our features is validated by experiments.

Key words: Spectral clustering,Shift invariant feature,Image feature,SAR image,Image patch

[1] Matas J,Churn O,et al.Robust wide baseline stereo from maximally stable external regions[J].Image and Vision Computing,2004,22(10):761-767
[2] 王国力,周伟,等.基于单演信号理论的SAR图像配准[J].电子与信息学报,2013,5(8):1779-1785
[3] 王永明,王贵锦.图像局部不变形特征与描述[M].北京:国防工业出版社,2010
[4] Candès E J,Demanet L,Donoho D L,et al.Fast discrete curvelet transforms[J].Applied and Computational Mathematics,California Institute of Technology,2005,5(3):861-899
[5] Candès E J,Donoho D L.Ridgelets:A key to higher-dimensionalintermittency?[J].Phil.Trans.R.Soc.A,1999(357):2495-2509
[6] Rosasco L,Belkin M,De Vito E.On learning with integral opera-tors[J].Journal of Machine Learning Research,2010(11):905-934
[7] von Luxberg U,Belkin M,Bousquet O.Consistency of spectral clustering[J].The Annals of Statistics,2008(36):555-586
[8] 韩彦彬.正定积分算子的本征值[J].科学通报,1986,0(17):1357-1357
[9] Fowlkes C,Belongie S,Chung F,et al.Spectral grouping using the Nystrm method[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2004,6(2):214-225
[10] Ng A Y,Jordan M I,Weiss Y.On spectral clustering:analysis and an algorithm[J].Advances in Neural Information Processing Systems,MIT Press,2001(14):849-856
[11] Mahoney M W,Orecchia L,Vishnoi N K.A local spectral methodfor graphs:with applications to improving graph partitions and exploring data graphs locally[J].Journal of Machine Learning Research,2012(13):2339-2365
[12] Maji S,Vishnoi N K,Malik J.Biased normalized cuts[C]∥Proceedings of the 2011 IEEE Conference on Computer Vision and Pattern Recognition.2011:2057-2064
[13] Belkin M,Rademacher L,Voss J R.The hidden convexity of spectral clustering.http://arxiv.org/abs/1403.0667
[14] Arias-Castro E,Lerman G,Zhang T.Spectral clustering based on local PCA.http://arxiv.org/abs/1301.2007
[15] Dhanjal C,Gaudelb R,Clémenonc R.Efficient eigen-updating for spectral graph clustering[J].Neurocomputing,2014,131:440-452
[16] Smith L M,Lerman K,Garcia-Cardona C,et al.Spectral clustering with epidemic diffusion[J].Physical Review E,2013,8(4):042813
[17] Mall R,Langone R,Suykens J A K.Kernel spectral clustering for big data networks[J].Entropy,2013(15):1567-1586
[18] 管涛,杨婷.谱聚类广义模型与典型算法分析[J].模式识别与人工智能,2014,27(11):1-11

No related articles found!
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!