A novel remote sensing classification rule extraction method based on discrete rough set

Land cover information which has been identified as the crucial data for land use planning and management have important economic value. In order to obtain land cover information, utilizing computer simulation technology to the automatic classify the remote sensing images is a very effective instrument. Rough set theory in dealing with remote sensing image's uncertainty, inconsistency and feature selection has a lot of advantages. However, the existing rough set methods is too sensitive to the spectral confusion between-class and spectral variation within-class, especially the classification rules extract by rough set may lead to the over-fitting phenomenon in the simulation process; this would limit the classification ability of rough sets. According to this case, this paper proposed a novel classification method based on rough set theory, improved the rules matching mechanism. Simulation results show that this method can reduce over-fitting phenomenon and the classification accuracy was improved.