Classification of Buildings and Roads Using Support Vector Machine

In this paper, it is intended to accurately separate pixels related to two spectrally similar classes of building and road in Shiraz urban area. To achieve this goal, Support Vector Machine (SVM) classification algorithm has been applied to a Landsat ETM+ image of Shiraz City. In order to assess the accuracy of the results, Maximum Likelihood Classification (MLC) as an approved and conventional algorithm has been applied on the image too. A visual and numerical comparison between these classification methods is carried out. Numerical comparison has been performed through overall accuracy and kappa coefficient applied on confusion matrices. From the assessment, it can be concluded that SVM classification method yields better results in the separation of pixels, especially on those related to two spectrally similar classes of building and road in this urban area.