Issues in training SVM classifications

SVM classification has great potential in remote sensing. The nature of SVM classification also provides opportunities for accurate classification from relatively small training sets, especially if interest is focused on a single class. Five approaches to reducing training set size from that suggested by conventional heuristics are discussed: intelligent selection of the most informative training samples, selective class exclusion, acceptance of imprecise descriptions for spectrally distinct classes, the adoption of a one-class classifier and a focus on boundary regions. All five approaches were able to reduce the training set size required considerably below that suggested by conventional widely used heuristics without significant impact on the accuracy with which the class of interest was classified. For example, reductions in training set size of ~90% from that suggested by a conventional heuristic are reported with the accuracy of the class of interest remaining nearly constant at ~95% and ~97% from the user's and producer's perspectives respectively.

[1]  Steven E. Franklin,et al.  A three-stage classifier for remote sensing of mountain environments , 1992 .

[2]  G. Foody.  Thematic map comparison: Evaluating the statistical significance of differences in classification accuracy , 2004 .

[3]  Vladimir N. Vapnik,et al.  The Nature of Statistical Learning Theory , 2000, Statistics for Engineering and Information Science.

[4]  Andreas Christmann,et al.  Support vector machines , 2008, Data Mining and Knowledge Discovery Handbook.

[5]  R. M. Lark,et al.  Components of accuracy of maps with special reference to discriminant analysis on remote sensor data , 1995 .

[6]  Gabriele Moser,et al.  Partially Supervised classification of remote sensing images through SVM-based probability density estimation , 2005, IEEE Transactions on Geoscience and Remote Sensing.

[7]  S. Ustin,et al.  Mapping nonnative plants using hyperspectral imagery , 2003 .

[8]  Chih-Jen Lin,et al.  A comparison of methods for multiclass support vector machines , 2002, IEEE Trans. Neural Networks.

[9]  David A. Landgrebe,et al.  Robust parameter estimation for mixture model , 1998, IGARSS '98. Sensing and Managing the Environment. 1998 IEEE International Geoscience and Remote Sensing. Symposium Proceedings. (Cat. No.98CH36174).

[10]  R. Nelson,et al.  Classifying northern forests using Thematic Mapper Simulator data , 1984 .

[11]  Russell G. Congalton,et al.  Assessing the accuracy of remotely sensed data : principles and practices , 1998 .

[12]  Martin Brown,et al.  Linear spectral mixture models and support vector machines for remote sensing , 2000, IEEE Trans. Geosci. Remote. Sens..

[13]  Giles M. Foody,et al.  Training set size requirements for the classification of a specific class , 2006 .

[14]  L. S. Davis,et al.  An assessment of support vector machines for land cover classi(cid:142) cation , 2002 .

[15]  Jim Piper,et al.  Variability and bias in experimentally measured classifier error rates , 1992, Pattern Recognit. Lett..

[16]  Giles M. Foody,et al.  Toward intelligent training of supervised image classifications: directing training data acquisition for SVM classification , 2004 .

[17]  Lorenzo Bruzzone,et al.  Classification of hyperspectral remote sensing images with support vector machines , 2004, IEEE Transactions on Geoscience and Remote Sensing.

[18]  Robert P. W. Duin,et al.  Support Vector Data Description , 2004, Machine Learning.

[19]  A. Belousov,et al.  A flexible classification approach with optimal generalisation performance: support vector machines , 2002 .

[20]  Giles M. Foody,et al.  Mapping a specific class for priority habitats monitoring from satellite sensor data , 2006 .

[21]  Lorenzo Bruzzone,et al.  A semilabeled-sample-driven bagging technique for ill-posed classification problems , 2005, IEEE Geoscience and Remote Sensing Letters.

[22]  Giles M. Foody,et al.  The use of small training sets containing mixed pixels for accurate hard image classification: Training on mixed spectral responses for classification by a SVM , 2006 .

[23]  B. Datt,et al.  On the relationship between training sample size and data dimensionality: Monte Carlo analysis of broadband multi-temporal classification , 2005 .

[24]  Clarence M. Sakamoto,et al.  LACIE - an application of meteorology for United States and foreign wheat assessment. , 1980 .