Hyperspectral images classification based on MKSVM and MRF
TAN Xiong1,2, YU Xuchu1, ZHANG Pengqiang1, FU Qiongying1, WEI Xiangpo1, GAO Meng1
1. Information Engineering University, Zhengzhou 450001, China; 2. Jiangxi Province Key Lab for Digital Land, East China Institute of Technology, Nanchang 330000, China
摘要为充分利用高光谱遥感影像中丰富的光谱和空间信息,提出了一种基于多核支持向量机(multiple kernel support vector machine,MKSVM)和马尔科夫随机场(markov random field,MRF)的影像分类方法。该方法首先利用MKSVM分类器对影像进行分类处理,再利用MRF对初始分类结果进行空间结构规则化,得到最终分类结果。通过对AVIRIS高光谱影像的分类实验表明,该方法有效地消除了分类结果中同质区域内的"噪声",分类精度提高了3%左右。
Abstract：To fully utilize the spectral and spatial information rich in hyperspectral remote sensing images, this paper proposes a hyperspectral images classification method based on multiple kernel support vector machine (MKSVM) and Markov random field (MRF). Firstly, the MKSVM classifier is used to classify hyperspectral images, then the MRF is used to regularize the initial classification results in the spatial structure, and the final classification results are obtained in the end. The experiment on AVIRIS hyperspectral image shows that the proposed method not only effectively eliminates the "noise" in the homogeneous regions within the classification results but also improves the classification accuracy by about 3%.
[1] 余旭初,冯伍法,林丽霞.高光谱——遥感测绘的新机遇[J].测绘科学技术学报,2006,23(2):101-105. Yu X C,Feng W F,Lin L X.Hyperspectralmote sensing:A new opportunity for surveying and mapping[J].Journal of Geomatics Science and Technology,2006,23(2):101-105. [2] Fauvel M,Benediktsson J A,Chanussot J.Spectral and spatial classification of hyperspectral data using SVMs and morphological profiles[J].IEEE Transactions on Geoscience and Remote Sensing,2008,46(11):3804-3814. [3] 吴昊.综合纹理特征的高光谱遥感图像分类方法[J].计算机工程与设计,2012,33(5):1993-1996. Wu H.Classification methodology combined with texture feature for hyperspectral remote sensing image[J].Computer Engineering and Design,2012,33(5):1993-1996. [4] Li C H,Kuo B C,Lin C T,et al.A spatial-contextual support vector machine for remotely sensed image classification[J].IEEE Transactions on Geoscience and Remote Sensing,2012,50(3):784-799. [5] Tarabalka Y,Tilton J C,Benediktsson J A,et al.Marker-based Hierarchical Segmentation and Classification Approach for Hyperspectral Imagery[C]//ICASSP'11.Prague Congress Center.Prague,Czech Republic,2011:1089-1092. [6] Tarabalka Y,Fauvel M,Chanussot J,et al.SVM and MRF based method for accurate classification of hyperspectral images[J].IEEE Transactions on Geoscience and Remote Sensing Letters,2010,7(4):736-740. [7] Zhang B,Li S S,Jia X P,et al.Adaptive Markov random field approach for classification of hyperspectral imagery[J].IEEE Transactions on Geoscience and Remote Sensing Letters,2011,8(5):973-977. [8] 杨国鹏,余旭初,周欣,等.基于相关向量机的高光谱影像分类研究[J].测绘学报,2010,39(6):572-578. Yang G P,Yu X C,Zhou X,et al.Research on relevance vector machine for hyperspectral imagery classification[J].Acta Geodaetica et Cartographica Sinica,2010,39(6):572-578. [9] 汪洪桥,孙富春,蔡艳宁,等.多核学习方法[J].自动化学报,2010,36(8):1037-1050. Wang H Q,Sun F C,Cai Y N,et al.On multiple kernel learning methods[J].Aata Automatica Sinica,2010,36(8):1037-1050. [10] 尚坤,李培军,程涛.基于合成核支持向量机的高光谱土地覆盖分类[J].北京大学学报:自然科学版,2011,47(1):109-114. Shang K,Li P J,Cheng T.Land cover classification of hyperspectral data using composite kernel support vector machines[J].Acta Scientiarum Naturalium Universitatis Pekinensis,2011,47(1):109-114. [11] Gu Y F,Wang C,You D,et al.Representative multiple kernel learning for classification in hyperspectral imagery[J].IEEE Transactions on Geoscience and Remote Sensing,2012,50(7):2852-2865. [12] 谭熊,余旭初,秦进春,等.高光谱影像的多核SVM分类[J].仪器仪表学报,2014,35(2):405-411. Tan X,Yu X C,Qin J C.et al.Multiple kernel SVM classification for hyperspectral images[J].Chinese Journal of Scientific Instrument,2014,35(2):405-411. [13] Platt J C.Probabilistic Outputs for Support Vector Machines and Comparisons to Regularized Likelihood Methods[OL].(1999).http://research.miscrosoft.com/~jplatt.
2015, Vol. 27 
