Support vector random field based approach towards object based classification of remotely sensed imagery

摘要

Remote sensing techniques are widely used for land cover classification and related analyses; however the availability of high resolution images have limited the accuracy of pixel based approaches. In this paper, we have analyzed the feasibility of incorporating contextual information to a support machine and have evaluated its performances with reference to the traditional approaches. Accuracy improvement of the proposed approach may be attributed to the effectiveness in combining spatial and spectral information.