摘要针对无法直接获取训练样本的遥感影像分类问题,从满足条件的其他影像中选择替代训练样本是最直接的方法,但由于地物类型在不同影像中的辐射环境不同,导致替代训练样本对待分类影像的代表性较差,无法保证分类精度。以直推式支持向量机(transductive support vector machine,TSVM)分类为例,发展了一种基于半监督学习的遥感影像训练样本时空拓展方法。该方法采用非监督方法从待分类影像中选择大量未标记样本,挖掘各类地物在特征空间中的结构信息; 以替代训练样本所拟合的分类面为初始面,通过自适应渐进式的优化,实现对待分类影像的高精度分类。该方法要求训练样本的来源影像与待分类影像具有相似的地物分布和相近的时相。以SPOT5和QuickBird影像分类为例,分别通过基于像元的和基于分割对象的分类实验证实,该文提出的方法可有效地实现训练样本的时空拓展应用。
Abstract:In classification of remote sensing images without any training samples, the choice of training samples from other representative images might be the only direct way; nevertheless, due to the difference of radiometric environments, the classification training samples from one image could not be well representative of other images. It is known that labeled samples from one image may not be effective for classifying others with high accuracy. In view of the above problem, a novel semi-supervised transcductive support vector machine(TSVM)method is proposed. The authors first chose a large quantities of unlabeled samples from the images which need to be classified in an unsupervised way, then extracted the inherent construction information of different classes in the feature space. Next, with the help of semi-supervised learning theory, the authors trained a classifier which was pre-trained by the labeled samples from another image in a recursive way, and at last an optimized classifier was obtained. It should be noted that two images involved in the method must have familiar land covers and acquired times. Classification experiments of SPOT5 and QuickBird remote sensing images were undertaken by the authors, and the classification results prove that the method proposed in this paper can effectively realize the sample extending application in both time and spatial domain.
任广波, 张杰, 马毅, 宋平舰. 基于半监督学习的遥感影像分类训练样本时空拓展方法[J]. 国土资源遥感, 2013, 25(2): 87-94.
REN Guangbo, ZHANG Jie, MA Yi, SONG Pingjian. Extending method of remote sensing image training sample based on semi-supervised learning in both time and spatial domain. REMOTE SENSING FOR LAND & RESOURCES, 2013, 25(2): 87-94.
[1] Jackson Q,Landgrebe D A.An adaptive classifier design for high-dimensional data analysis with a limited training data set[J].IEEE Transactions on Geoscience and Remote Sensing,2001,39(12):2664-2679. [2] Jensen J R.Introductory digital image processing:A remote sensing perspective[M].3rd ed.New Jersey:Prentice Hall,2006:195-209. [3] Kaufman Y J.Fraster R S.Atmospheric effect on classification of finite fields[J].Remote Sensing of Environment,1984,15(2):95-118. [4] Cracknell A P,Hayes L W.Atmospheric corrections to passive satellite remote sensing data[M]//Cracknell A P,Hayes L W B.Chapter 8 in Introduction to Remote Sensing.London:Taylor and Francis,1993:116-158. [5] Camps V G,Bandos T,Zhou D Y.Semi-supervised graph based hyperspectral image classification[J].IEEE Transaction on Geoscience and Remote Sensing,2007,45(10):3044-3054. [6] Du Y,Teillet P M,Cihlar J.Radiometric normalization of multitemporal high-resolution satellite images with quality control for land cover change detection[J].Remote Sensing of Environment,2002,82(1):123-134. [7] 张友水, 冯学智,周成虎.多时相TM影像相对辐射校正研究[J].测绘学报,2006,35(2):122-127. Zhang Y S,Feng X Z,Zhou C H.Relative radiometric correction for multitemporal TM images[J].Acta Geodaetica et Cartographica Sinica,2006,35(2):122-127. [8] Koukal T,Suppan F,Schneider W.The impact of relative radiometric calibration on the accuracy of KNN-predictions of forest attributes[J].Remote Sensing of Environment,2007,110(4):431-437. [9] Ren G B,Zhang J,Ma Y,et al.A method for classification training sample spatial-time expanding of remote sensing images[C]//International Conference on Space Information Technology,Beijing,Proceedings of the SPIE,2009:76510G1-G7. [10] Scudder I I.Probability of error of some adaptive pattern-recognition machines[J].IEEE Transactions on Information Theory,1965,11(3):363-371. [11] Fralick S.Learning to recognize patterns without a teacher[J].IEEE Transactions on Information Theory,1967,13(1):57-64. [12] Nigam K,McCallum A,Mitchell T.Semi-supervised text classification using EM semi-supervised learning[M].Cambridge MA:MIT Press,2006:33-55. [13] István N T,Richárd F,János C.On positive and unlabeled learning for text classification[C]//István N T,Richárd F,János C.Lecture Notes in Computer Science.London:Springer,2011,6836:219-226. [14] Schenker A,Bunke H,Last M,et al.A graph-based framework for web document mining[M]//Schenker A,Bunke H,Last M,et al.Document Analysis Systems.London:Springer,2004:401-412. [15] Yang Y,Wu F,Nie F,et al.Web and personal image annotation by mining label correlation with relaxed visual graph embedding image processing[J].IEEE Transactions,2012,21(3):1339-1351. [16] Shahshahani B M,Landgrebe D A.The effect of unlabeled samples in reducing the small sample size problem and mitigating the hughes phenomenon[J].IEEE Transactions on Geoscience and Remote Sensing,1994,32(5):1087-1092. [17] 骆剑承,王钦敏,马江洪,等.遥感图像最大似然分类方法的EM改进算法[J].测绘学报,2002,31(3):234-239. Luo J C,Wang Q M,Ma J H,et al.The EM-based maximum likelihood classifier for remotely sensed data[J].Acta Geodaetica et Cartographica Sinica,2002,31(3):234-239. [18] Tuia D,Camps V G.Semi-supervised remote sensing image classification with cluster kernels[J].IEEE Geoscience and Remote Sensing Letters,2009,6(2):224-228. [19] 任广波,张杰,马毅,等.生成模型学习的遥感影像半监督分类[J].遥感学报,2010,14(6):1090-1104. Ren G B,Zhang J,Ma Y,et al.Generative model based semi-supervised learning method of remote sensing image classification[J].Journal of Remote Sensing,2010,14(6):1090-1104. [20] Tuia D,Pasolli E,Emery W J. Using active learning to adapt remote sensing image classifiers[J].Remote Sensing of Environment,2011,115(9):2232-2242. [21] Galante N,Siqueia R,Sant’Anna S J,et al.Semi-supervised remote sensing image classification methods assessment[C]//IGARSS. Geoscience and Remote Sensing Symposium.Vancouve:IEEE International,2011:2939-2942. [22] Mishra N S,Ghosh S,Ghosh A.Semi-supervised fuzzy clustering algorithms for change detection in remote sensing images[C]//Mishra N S,Ghosh S,Ghosh A.Lecture Notes in Computer Science.London:Springer,2012,7143:269-276. [23] Joachims T.Transductive inference for text classification using support vector machines[C]//Proceedings of the 16th International Conference on Machine Learning(ICML).San Francisco:Morgan Kaufmann Publishers,1999:200-209. [24] 陈毅松,汪国平,董士海.基于支持向量机的渐进直推式分类学习算法[J].软件学报,2003,14(3):451-460. Chen Y S,Wang G P,Dong S H.A progressive transductive inference algorithm based on support vector machine[J].Journal of Software,2003,14(3):451-460. [25] 沈新宇,许宏丽,官腾飞.基于直推式支持向量机的图像分类算法[J].计算机应用,2007,27(6):1463-1464. Shen X Y,Xu H L,Guan T F.Image classification based on transductive support vector machines[J].Computer Application,2007,27(6):1463-1464. [26] 廖东平,魏玺章,黎湘,等.一种改进的渐进直推式支持向量机分类学习算法[J].信号处理,2008,24(2):213-218. Liao D P,Wei X Z,Li X,et al.An improved learning algorithm with progressive transductive support vector machine[J].Signal Processing,2008,24(2):213-218.