Accuracy enhancement in land use/cover classification based on extractable indices of variorum in satellite images

In this study texture and variance information are used in some cases but extraction of data from variorum is less carried out and its time changes has not been noticed enough. Multitemporal satellite images of SPOT and ASTER are used in this study for mapping land use/cover of Dehgolan in Kordestan region. The extracted set of indices from variorum is introduced and a new series of indices was extracted and classified by using SVM method. The whole process was accomplished in two stages: 1) classification of spectral band, 2) classification of spectral band along with the parameters extracted from the variorum. After accuracy assessment and their comparison, appropriate indices were suggested. Some indices increase the accuracy of most classes while may have no impact on other classes; these can be considered public indices. Moreover, some indices which are called included specialty indices increase the accuracy of some classes and reduce the accuracy of others. The performed studies, demonstrated that using the indices derived from variorum increases the overall accuracy by 4 to 9 percent in some classes and the overall accuracy is enhanced from 88.98% to 92.744%.