摘要
针对目前常用的基于像素的深度神经网络极化SAR分类方法产生的椒盐现象,文中提出了一种联合自适应阈值多尺度分割方法和径向基神经网络的极化SAR地物分类方法。实验证明,该方法能够有效地保留SAR图像的结构特征并有效消除分类过程中产生的椒盐现象和破碎斑块,具有较高的分类精度。
Pixel-based deep neural network polarization SAR classification method will produce salt and pepper phenomenon. In order to solve this kind of phenomenon,this paper proposes a method of classification of polarimetric SAR by combining the adaptive threshold multi-scale segmentation method and RBF neural network. Experimental results show that this method can effectively preserve the structural features of SAR images and effectively eliminate salt and pepper phenomena and broken plaque generated during classification. Therefore,this algorithm has higher classification accuracy.
引文
[1]赵一博,邹焕新,秦先祥.一种基于RBF神经网络的极化SAR图像分类方法[J].现代雷达,2013,35(8):24-27.
[2]韩佳敏.基于深度RBF网络的SAR影像地物分类[D].西安:西安电子科技大学,2014.
[3] Orr M J L. Introduction to radial basis function networks[J]. Internationale Zeitschrift Für Vitaminforschung.interntional Journal of Vitamin Research. journal International De Vitaminologie,2003,37(3):2 797-2 800.
[4] B. Yekkehkhany,A. Safari,S. Homayouni,et al. A Comparison Study of Different Kernel Functions for SVMBASED Classification of Multi-temporal Polarimetry SAR DATA[J]. ISPRS-International Archives of the Photogrammetry,Remote Sensing and Spatial Information Sciences,2014,XL-2/W3(1):.
[5] Hamed Kashi,Samad Emamgholizadeh,Hadi Ghorbani.Estimation of Soil Infiltration and Cation Exchange Capacity Based on Multiple Regression, ANN(RBF,MLP),and ANFIS Models[J]. Communications in Soil Science and Plant Analysis,2014,45(9):.
[6] Abhisek Paul,Paritosh Bhattacharya,Santi Prasad Maity.Comparative Analysis of Radial Basis Functions with SAR Images in Artificial Neural Network[M].Springer International Publishing.
[7] Libby V. Feature-based classification of SAR data using RBF networks[J]. Proceedings of SPIE-The International Society for Optical Engineering,1995:583-594.
[8] Ince T,Kiranyaz S,Gabbouj M. Evolutionary RBF classifier for polarimetric SAR images[J]. Expert Systems with Applications,2012,39(5):4 710-4 717.
[9] Ince T. Polarimetric SAR Image Classification Using Radial Basis Function Neural Network[J]. Piers Online,2010,6(6):470-475.
[10] Paul A,Bhattacharya P,Maity S P. Comparative Analysis of Radial Basis Functions with SAR Images in Artificial Neural Network[J]. Advances in Intelligent Systems&Computing,2015(320):125-131.
[11] Yang J,Yang R. Unsupervised classification of polarimetric SAR images using complex Wishart distribution based on H/αdecomposition and algorithm evaluation[J]. 2007,6790.
[12] Yang,S. Y.,M. Wang,L. C. Jiao.Radar target recognition using contourlet packettransform and neural network approach[J].Signal Processing,2009,89(4):394-409.
[13] Zou T,Yang W,Dai D,et al. Polarimetric SAR Image Classification Using Multifeatures Combination and Extremely Randomized Clustering Forests[J]. Eurasip Journal on Advances in Signal Processing,2009,2010(1):1-9.
[14] Cao F,Hong W,Pottier E. An improvement for the unsupervised Wishart Freeman classification with fully polarimetric SAR data[C]//Geoscience and Remote Sensing Symposium. IEEE,2011:320-322.
[15] Johnson B,Xie Z. Unsupervised image segmentation evaluation and refinement using a multi-scale approach[J].Isprs Journal of Photogrammetry&Remote Sensing,2011,66(4):473-483.
[16] Zhao Q,Li Y,Liu Z. SAR image segmentation using Voronoi tessellation and Bayesian inference applied to dark spot feature extraction.[J]. Sensors,2013,13(11):14484-14 499.