Comparison between Object-based and Pixel-based Classification Methods upon High Resolution Quickbird Image

Abstract

Special spatial analysis models must be increasingly used for municipal land-use/land-cover organisation, but the high phantom change within the identical land-cover, the phantom jam with another land-covers, and the darkness difficulty frequently lead to lower order fulfillment based on the popular per-pixel spectral-based analysis plans. The study examines ways to increase urban land-cover kind with Quick Bird description.

Maximum likelihood pixel based supervised as well as Rule based object based approaches were examined on Quick bird satellite image in Karbala. The study shows that treatment of textural data through the rule-based model analysis method, can significantly improve land use classification performance. Moreover, the results show higher overall accuracy in the rule based method than pixel based. Particularly in urban extractions the object based performed much more capabilities.