摘要
合成孔径雷达(Synthetic Aperture Radar,SAR)图像分割的目的是从复杂的地物场景中提取和识别特定的目标。基于马尔可夫随机场(Markov Random field, MRF)模型的图像分割方法能够将标记图像的上下文信息和待分割图像的统计特性统一起来,特别适合SAR图像的分割,因此开展相关研究工作十分有必要。
论文以马尔可夫随机场理论为研究基础,系统归纳和总结了国内外研究成果和经验,按照SAR图像分割方法的基本步骤展开。论文的主要工作有:
(1)论述了在SAR图像分割中引入MRF模型的优点,并阐述了MRF模型的原理和方法;
(2)在参阅大量文献的基础上,总结了各种已有的杂波统计模型。重点介绍了乘积模型中的K分布模型和G~0分布模型;
(3)在总结杂波统计模型的基础上,给出了各种参数模型的参数估计方法。着重介绍了G~0分布的参数估计方法,并在已有方法的基础上提出了一种Mellin变换和矩估计相结合参数估计方法;
(4)总结了各类MRF模型求最优解的迭代算法,对各种迭代算法的特点进行了分析,并对其中的典型算法进行仿真比较;
(5)通过优化MRF模型的邻域结构,引入了一种具有图像细节保护能力的自适应邻域结构的SAR图像分割方法,并利用合成SAR图像和真实SAR图像与传统MRF模型的分割结果进行对比分析,结果表明,该分割方法可以较好的保留图像细节。
The aim of Synthetic Aperture Radar (SAR) images segmentation is extracting and detecting specific targets from complex ground scene. The segmentation methods of images based on Markov Random Field (MRF) model able to unify the contextual information from label images and statistical properties of observed images, which are especially suitable for segmentation of SAR images. Therefore, developing a relevant research work on this field is necessary.
In this paper, we start our works according to the basic steps of SAR images segmentation based on MRF theory, also concluded and summarized research works and experiences at home and abroad. The work of this paper can be concluded as follows:
(1) Discussed the advantages that introduce MRF model to SAR images segmentation, and described the principles and methods of MRF model;
(2) Summarized all kinds of existing clutter statistical models based on reading extensive literatures, K distribution model and G~0 distribution model are our focuses;
(3) Summarized parameter estimation methods of existing clutter statistical models, introduced G~0 distribution model parameter estimation methods emphatically, then proposed a new parameter estimation method which combines the methods of moments and the method based on Mellin transform on the basis of existing G~0 distribution model parameter estimation methods.
(4) Summarized the various types of existing iterative algorithms of MRF model, and analyzed the characteristics of existing iterative algorithms, then some typical algorithms are simulated and compared;
(5) Introduced a new SAR images segmentation method with details preserving ability based on adaptive neighborhoods by optimizing the neighborhoods structure of MRF model. We use synthetic and real SAR images to compare the segmentation results, the results show that the segmentation method can preserve image details better.
引文
[1] Wiley C A.Synthetic aperture radars-a paradigm for technology evolution [J]. IEEE Trans. Aerospace Elec. Sys.1985, AES- 21:440-443.
[2]张直中.机载和星载合成孔径雷达导论[M].北京:电子工业出版社,2004
[3]保铮,邢孟道,王彤.雷达成像技术[M].北京:电子工业出版社,2005
[4] Kirk J C,Jr.Digital synthetic aperture radar technology[A].IEEE International Radar Conference Record[C] .1975:482～487
[5] Bennett J R , Widmer P , Cumming I G . A real-time airborne SAR processor[J].Proc. ESA Working Group,Erascati,Italy,1980,ESC SP-1031.
[6]匡纲要,高贵,蒋咏梅.合成孔径雷达目标检测理论、算法及应用[M].长沙:国防科技大学出版社,2007
[7] Rogers S K Ruck,D W Tart,G L Kabrisky.Synthetic aperture radar segmentation using wavelets and fractals[A].Systems Engineering[C].IEEE ICPD1991,1991(8):21-24
[8] Y Dong,B C Forster.Segmentation of radar imagery using Gaussian Markov random field models and wavelet and transform technique[C] . IGARSS 97.1997.2054-2056
[9] V Venkatachalam,H Choi.Multiscal SAR image segmentation using wavelet domain hidden Markov tree model[J].SPIE 1998,3:141-151
[10] Juha A Karvonen.Baltic sea ice SAR segmentation and classification using modified pulse-coupled neural networks[J].IEEE Trans GARS,2004,42(7):1566-1574
[11] H Derin,P Kelly.Modeling and segmentation of speckled images using complex data[J].IEEE Trans on Geosci Remote Sensing,1990,28(1):76-87
[12] G Davidson,K Ouchi.Segmentation of SAR images using multitemporal information[J].Radar,Sonar and Navigation,IEEE Proceedings,2003,150(10):367-374
[13] Quelle H C,Delignon Y,Marzouki A.Unsupervised Bayesian segmentation of SAR images using the Pearson system distributions[C].IGARSS 93,1993,3(9):1538-1540
[14] Tison C,Nicolas J M,Tupin F,et al.A new statistical model for Markovian classification of urban areas in high—resolution SAR images.IEEE Trans on GRS,2004,42(10):2046-2057
[15]夏桂松,何楚,孙洪.一种基于非参数密度估计和马尔可夫上下文的SAR图像分割算法[A].电子与信息学报.2006.12.Vol.28.No.12
[16] Abend,T J Harley,L N Kanal.Classification of binary random patterns[J].IEEE Trans Inform Theory IT-11,1965:538-544
[17] J M Hammersley , P Clifford . Markov field on finite graphs and lattices.unpublished,1971
[18] J Besag.Spatial interaction and the statistical analysis of lattice systems.Journal of the Royal Statistical Society B,1974,36:192–236
[19] S Geman,D Geman.Stochastic Relaxation, Gibbs Distribution, and the Bayesian Restoration of Images[J] .IEEE Trans Pattern Analysis Machine Intelligence,1984,PAMI-6721-741
[20]曹建农,王占宏.图像分割中MRF与DMN方法评述与比较[J].计算机工程与应用,2006,20:19-24
[21]朱国普,曾庆双,屈彦呈等.一种基于HMRF模型的无监督图像分割算法[J] .电子学报,2006,Vol.34,No.2:374-379
[22]汪西莉,焦李成.基于多尺度马尔可夫随机场的图像分割[J].计算机科学,2003,Vol.30,No.7:174-176
[23]陆明俊,王润生.基于MRF模型的可靠的图像分割[J].电子学报,1999,Vol.27,No.2:87-89
[24]钟平.面向图像标记的随机场模型研究.国防科技大学博士学位论文,2008.10.
[25] Oliver C J.Understanding Synthetic Aperture Radar Images[M].Boston/London:Artech House,1998
[26] Arseoault H,et al.Properties of Speckle Integrated with a Finite Aperture and Logarithmically Transformed [J].J.Opt.Soc.Am.,1976,66(11):1160-1163.
[27] Ward K D.Compound Representation of High Resolution Sea Clutter[J].Electron Lett.,1981,7:561-565
[28]高贵.SAR图像统计建模研究综述[A].信号处理.2009.8.Vol.23.No.8:1270-1278
[29]邹焕新.SAR图像舰船目标与航迹检测方法研究[D].国防科技大学博士学位论文,2003
[30] Jiang Q , et al . Ship Detection in RADARSAT SAR Imagery Using PNN-model[C].Proceedings of ADRO symposium’98.1998,5:4562-4566
[31] Bishop C M.Neural Networks for Pattern Recognition[M].2nd Edition,Oxford University Press,1996
[32] Mantero P,et al.Partially Supervised Classification of Remote Sensing Images Using SVM-based Probability Density Estimation [M] . IEEE Honorary Workshop.2003
[33] Vapnik V N.Statistical Learning Theory[M].John Wiley and Sons Inc..1998.
[34] George S W . The Detection of Nonfluctuating Targets in Log-normal Clutter[R].NRL Report,1968,6796
[35] Kuruoglu E E,et al.Modeling SAR Images With a Generalization of the Rayleigh Distribution[J].IEEE Trans.Image Processing,2004,13(4):191-197
[36] Ulaby F T,et al.Textural Information in SAR Images[J].IEEE Trans.GRS,1986,24:235-245
[37] Frery A C, Muller H J, Yanasse C F, et a1.A model for extremely heterogeneous clutter.IEEE Trans on GRS.1997, 35(3):648-659
[38] Freitas C C,Frery A C,Correia A H.The Polari metric G distribution for SAR data analysis.Environmetries,2005,16(1):13-31
[39]时松涛,高贵等.基于Mellin变换的G0分布参数估计方法.自然科学进展. 19(6) :677-689
[40]高贵.SAR图像目标ROI自动获取技术研究.长沙:国防科技大学博士学位论文,2007
[41] Delignon Y , et al . Modeling Non-Rayleigh Speckle Distribution in SAR Images[J].IEEE Trans.GRS,2002,40(6):1430-1435
[42] Anastassopoulos V,et al.High Resolution Radar Clutter Statistics[J].IEEE Trans.AES,1999,35(1):43-59
[43] Pierce R D.RCS Characterization Using the Alpha-stable Distribution and Applications [C].IEEE 1996 Nat. Radar Conf.,1996:394-419
[44] Goodman J W.Statistical Properties of Laser Speckle Patterns, Laser Speckle and Related Phenomena [M].J.C. Dinty, Ed. New York:Academic,1980.
[45] Blacknell D.A Mixture Distribution Model for Correlated SAR Clutter[C].SPIE,1996,2958:38-49
[46] Blacknell D.Target Detection in Correlated SAR Clutter [J].IEEE Proc-RSN,2000,147(1):9-16
[47] Blake A P , et al . High Resolution SAR Clutter Textural Analysis and Simulation.SPIE,1995,2584:101-108
[48] Blake A P,et al.High Resolution SAR Clutter Textural Analysis[C].IEEE Colloquium on Recent Developments in Radar and Sonar Imaging Systems:What Next? 1995,10:1-9
[49]张琦,等.多纹理高分辨率SAR图像杂波统计模型[C].第二届中国合成孔径雷达会议论文集,2005:327-331
[50] Akaike H.Information Theory and an Extension of Maximum Likelihood Principle [C].2nd International Symposium on Information theory,1973:261-281
[51] DcVore M D,et a1.Statistical Assemmment of Model Fit for Synthetic Aperture Radar Data[C].SPIE,2001,4382:379-388
[52] Besag J.On the Statistical Analysis of Dirty Pictures[J].Journal of Royal Statistical Society, 1986, B48:259-302
[53] Metropolis N A,A Rosenbluth,M Rosenbluth,et al.Equation of state calculations by fast computing machines.Journal of Chemical Physics,vol.21,1953:1087-1092
[54] Kirkpatrick S,C D Gelatt Jr,P Bruckner.Complexity of machine scheduling problems.Annals of Discrete Mathematics,vol.7,1977:343-362
[55]陈国良,王煦法,庄镇泉.遗传算法及其应用[M].北京:人民邮电出版社。1996
[56] A P Dempster,N M Laird,D B Rubin.Maximum Likelihood from Incomplete Data via the EM Algorithm.Journal of the Royal Statistical Society,Series B,vol.39,no.1,1977:1-38
[57] S Z Li.Markov random field modeling in image analysis. Springer-Verlag, 2001.
[58] J N Provost, C Collet, P Rostaing, et al.Hierarchical Markovian Segmentation of Multispectral Images for the Reconstruction of Water Depth Maps. Computer Vision & Image Understanding, 2004, 93 (2):155-174
[59] Smits P C,Dellepiane S G.Synthetic aperture radar image segmentation by a detail preserving Markov Random Field approach[J].IEEE Transactions on Geosciences and Remote Sensing,1997,35(4):844-857