特征选择与深度学习相结合的极化SAR图像分类
|
摘要
给出了一种特征选择与深度学习相结合的极化合成孔径雷达(polarimetric synthetic aperture radar, PolSAR)图像有监督分类算法。该算法首先根据极化SAR图像数据以及目标分解获取原始特征参数集,然后利用随机森林(Random Forest, RF)方法对特征参数集进行重要性评估,并根据特征重要性排名选择最优极化特征。以最优极化特征为输入,通过卷积神经网络(convolutional neural network, CNN)学习多层特征信息,再利用训练好的网络模型对极化SAR图像进行分类。利用美国AIRSAR机载系统采集的实测数据进行实验,并同已有经典有监督分类算法进行比较,结果表明本文算法能够选取有效的极化特征,最终得到较为准确的分类效果。
A supervised classification algorithm with feature selection and deep learning for polarimetric synthetic aperture radar(PolSAR) image is proposed in this paper. Firstly, an original feature parameter set is extracted from the polarization SAR image data and the target decomposition. Then the random forest method is used to evaluate the importance of the feature parameter set. After that, the optimal polarization features are obtained according to the feature importance rank.Taking the optimal polarization feature as the input, the multi-layer feature information is learned by the convolutional neural network(CNN), and the PolSAR image is classified by the trained network model. Experiments are carried out using the measured data collected by the U.S. AIRSAR airborne system, and the results are compared with which of the existing classical supervised classification algorithm. The results show that the proposed algorithm can select effective polarization features and finally obtain more accurate classification results.
A supervised classification algorithm with feature selection and deep learning for polarimetric synthetic aperture radar(PolSAR) image is proposed in this paper. Firstly, an original feature parameter set is extracted from the polarization SAR image data and the target decomposition. Then the random forest method is used to evaluate the importance of the feature parameter set. After that, the optimal polarization features are obtained according to the feature importance rank.Taking the optimal polarization feature as the input, the multi-layer feature information is learned by the convolutional neural network(CNN), and the PolSAR image is classified by the trained network model. Experiments are carried out using the measured data collected by the U.S. AIRSAR airborne system, and the results are compared with which of the existing classical supervised classification algorithm. The results show that the proposed algorithm can select effective polarization features and finally obtain more accurate classification results.
引文
[1] 李仲森.极化雷达成像基础与应用[M].北京:电子工业出版社,2013.Lee J S.Polarimetric Radar Imaging:From Basics to Applications[M].Beijing:Publishing House of Electronics Industry,2013.(in Chinese)
[2] 孙勋,杨祥立,涂尚坦,等.结合特征选择和大尺度谱聚类的极化SAR图像非监督分类[J].信号处理,2016,32(6):684- 693.Sun X,Yang X L,Tu S T,et al.Unsupervised Classification of PolSAR Images by Combining Feature Selection and Large Scale Spectral Clustering[J].Journal of Signal Processing,2016,32(6):684- 693.(in Chinese)
[3] Ren B,Hou B,Zhao J,et al.Unsupervised Classification of Polarimetirc SAR Image Via Improved Manifold Regularized Low-Rank Representation With Multiple Features[J].IEEE Journal of Selected Topics in Applied Earth Observations & Remote Sensing,2017,PP(99):1-16.
[4] Lee J S,Grunes M R,Pottier E,et al.Unsupervised terrain classification preserving polarimetric scattering characteristics[J].IEEE Transactions on Geoscience & Remote Sensing,2004,42(4):722-731.
[5] 陈强,蒋咏梅,陆军,等.一种基于目标散射鉴别的POLSAR图像地物无监督分类新方法[J].电子学报,2011,39(3):613- 618.Chen Q,Jiang Y M,Lu J,et al.A New Scattering-Identification Based Unsupervised Terrain Classification for POLSAR Image[J].Acta Electronica Sinica,2011,39(3):613- 618.(in Chinese)
[6] Lee J S,Grunes M R,Kwok R.Classification of multi-look polarimetric SAR imagery based on complex Wishart distribution[J].International Journal of Remote Sensing,1994,15(11):2299-2311.
[7] Lee J S,Pottier E.Quantitative Comparison of Classification Capability_ Fully Polarimetric Versus Dual and Single-Polarization SAR[J].IEEE Transactions on Geoscience & Remote Sensing,2001,39(11):2343-2351.
[8] 杨真真,匡楠,范露,等.基于卷积神经网络的图像分类算法综述[J].信号处理,2018,34(12):1474-1489.Yang Z Z,Kuang N,Fan L,et al.Review of Image Classification Algorithms Based on Convolutional Neural Networks[J].Journal of Signal Processing,2018,34(12):1474-1489.(in Chinese)
[9] 陈书贞,解小会,杨郁池,等.利用多尺度卷积神经网络的图像超分辨率算法[J].信号处理,2018,34(9):1033-1044.Chen S Z,Xie X H,Yang Y C,et al.Image Super-Resolution Algorithm Based on Multi-Scale Convolution Neural Network[J].Journal of Signal Processing,2018,34(9):1033-1044.(in Chinese)
[10] Uhlmann S,Kiranyaz S.Integrating Color Features in Polarimetric SAR Image Classification[J].IEEE Transactions on Geoscience & Remote Sensing,2014,52(4):2197-2216.
[11] Liu H,Yang S,Gou S,et al.Polarimetric SAR Feature Extraction With Neighborhood Preservation-Based Deep Learning[J].IEEE Journal of Selected Topics in Applied Earth Observations & Remote Sensing,2016,PP(99):1-11.
[12] Chen S W,Tao C S.Multi-temporal PolSAR crops classification using polarimetric-feature-driven deep convolutional neural network[C]//International Workshop on Remote Sensing with Intelligent Processing.IEEE,2017:1- 4.
[13] 石俊飞,刘芳,林耀海,等.基于深度学习和层次语义模型的极化SAR分类[J].自动化学报,2017,43(2):215-226.Shi J F,Liu F,Lin Y H,et al.Polarimetric SAR Image Classification Based on Deep Learning and Hierarchical Semantic Model[J].Journal of Automatica Sinica,2017,43(2):215-226.(in Chinese)
[14] 徐丰,王海鹏,金亚秋.深度学习在SAR目标识别与地物分类中的应用[J].雷达学报,2017,6(2):136-148.Xu F,Wang H P,Jin Y Q.Deep Learning as Applied in SAR Target Recognition and Terrain Classification[J].Journal of Radars,2017,6(2):136-148.(in Chinese)
[15] Li J,Wang C,Wang S,et al.Classification of very high resolution SAR image based on convolutional neural network[C]//International Workshop on Remote Sensing with Intelligent Processing.IEEE,2017:1- 4.
[16] Huynen J R.Phenomenological Theory of Radar Targets[J].Electromagnetic Scattering,1978:653-712.
[17] Cloude S R,Pottier E.A Review of Target Decomposition Theorems in Radar Polarimetry[J].IEEE Transactions on Geoscience and Remote Sensing,1996,34(2):498-518.
[18] Yamaguchi Y.Four-Component Scattering Model for Polarimetric SAR Image Decomposition based on Asymmetric Covariance Matrix[J].Technical Report of Ieice Sane,2005,104(8):1699-1706.
[19] Genuer R,Poggi J M,Variable selection using random forest[J].Pattern Recognition Letters,2010,31(14):2225-2236.
[2] 孙勋,杨祥立,涂尚坦,等.结合特征选择和大尺度谱聚类的极化SAR图像非监督分类[J].信号处理,2016,32(6):684- 693.Sun X,Yang X L,Tu S T,et al.Unsupervised Classification of PolSAR Images by Combining Feature Selection and Large Scale Spectral Clustering[J].Journal of Signal Processing,2016,32(6):684- 693.(in Chinese)
[3] Ren B,Hou B,Zhao J,et al.Unsupervised Classification of Polarimetirc SAR Image Via Improved Manifold Regularized Low-Rank Representation With Multiple Features[J].IEEE Journal of Selected Topics in Applied Earth Observations & Remote Sensing,2017,PP(99):1-16.
[4] Lee J S,Grunes M R,Pottier E,et al.Unsupervised terrain classification preserving polarimetric scattering characteristics[J].IEEE Transactions on Geoscience & Remote Sensing,2004,42(4):722-731.
[5] 陈强,蒋咏梅,陆军,等.一种基于目标散射鉴别的POLSAR图像地物无监督分类新方法[J].电子学报,2011,39(3):613- 618.Chen Q,Jiang Y M,Lu J,et al.A New Scattering-Identification Based Unsupervised Terrain Classification for POLSAR Image[J].Acta Electronica Sinica,2011,39(3):613- 618.(in Chinese)
[6] Lee J S,Grunes M R,Kwok R.Classification of multi-look polarimetric SAR imagery based on complex Wishart distribution[J].International Journal of Remote Sensing,1994,15(11):2299-2311.
[7] Lee J S,Pottier E.Quantitative Comparison of Classification Capability_ Fully Polarimetric Versus Dual and Single-Polarization SAR[J].IEEE Transactions on Geoscience & Remote Sensing,2001,39(11):2343-2351.
[8] 杨真真,匡楠,范露,等.基于卷积神经网络的图像分类算法综述[J].信号处理,2018,34(12):1474-1489.Yang Z Z,Kuang N,Fan L,et al.Review of Image Classification Algorithms Based on Convolutional Neural Networks[J].Journal of Signal Processing,2018,34(12):1474-1489.(in Chinese)
[9] 陈书贞,解小会,杨郁池,等.利用多尺度卷积神经网络的图像超分辨率算法[J].信号处理,2018,34(9):1033-1044.Chen S Z,Xie X H,Yang Y C,et al.Image Super-Resolution Algorithm Based on Multi-Scale Convolution Neural Network[J].Journal of Signal Processing,2018,34(9):1033-1044.(in Chinese)
[10] Uhlmann S,Kiranyaz S.Integrating Color Features in Polarimetric SAR Image Classification[J].IEEE Transactions on Geoscience & Remote Sensing,2014,52(4):2197-2216.
[11] Liu H,Yang S,Gou S,et al.Polarimetric SAR Feature Extraction With Neighborhood Preservation-Based Deep Learning[J].IEEE Journal of Selected Topics in Applied Earth Observations & Remote Sensing,2016,PP(99):1-11.
[12] Chen S W,Tao C S.Multi-temporal PolSAR crops classification using polarimetric-feature-driven deep convolutional neural network[C]//International Workshop on Remote Sensing with Intelligent Processing.IEEE,2017:1- 4.
[13] 石俊飞,刘芳,林耀海,等.基于深度学习和层次语义模型的极化SAR分类[J].自动化学报,2017,43(2):215-226.Shi J F,Liu F,Lin Y H,et al.Polarimetric SAR Image Classification Based on Deep Learning and Hierarchical Semantic Model[J].Journal of Automatica Sinica,2017,43(2):215-226.(in Chinese)
[14] 徐丰,王海鹏,金亚秋.深度学习在SAR目标识别与地物分类中的应用[J].雷达学报,2017,6(2):136-148.Xu F,Wang H P,Jin Y Q.Deep Learning as Applied in SAR Target Recognition and Terrain Classification[J].Journal of Radars,2017,6(2):136-148.(in Chinese)
[15] Li J,Wang C,Wang S,et al.Classification of very high resolution SAR image based on convolutional neural network[C]//International Workshop on Remote Sensing with Intelligent Processing.IEEE,2017:1- 4.
[16] Huynen J R.Phenomenological Theory of Radar Targets[J].Electromagnetic Scattering,1978:653-712.
[17] Cloude S R,Pottier E.A Review of Target Decomposition Theorems in Radar Polarimetry[J].IEEE Transactions on Geoscience and Remote Sensing,1996,34(2):498-518.
[18] Yamaguchi Y.Four-Component Scattering Model for Polarimetric SAR Image Decomposition based on Asymmetric Covariance Matrix[J].Technical Report of Ieice Sane,2005,104(8):1699-1706.
[19] Genuer R,Poggi J M,Variable selection using random forest[J].Pattern Recognition Letters,2010,31(14):2225-2236.