Multisource fusion for land cover classification using support vector machines

Remote sensing data have proven to be an attractive source for extracting accurate land cover information. For a given application, information from an individual sensor may be incomplete, inconsistent, and imprecise. Additional data sources may assist in achieving a higher degree of accuracy. Recently, support vector machines (SVM), a non-parametric algorithm, has been proposed as an alternative for classification of remote sensing data, and the results are promising. In this paper, the use of the SVM algorithm for multisource classification has been investigated. An IRS-1C LISS III image along with NDVI and DEM data layers in the Himalayan region were fused for classification. The results illustrate a significant improvement in accuracy of classification on incorporation of ancillary data over the classification performed solely on the basis of remote sensing data.