Zhang, Y.; Liu, J.; Wan, L., and Qi, S., 2015. Land cover/use classification based on feature selection.

Based on Support Vector Machine (SVM) and decision trees (DTs) classification methods, in this paper, we take Harbin, Daqing, Suihua, and Qiqihar in Heilongjiang Province as study area for land cover/use classification. Feature bands representing land surface characteristics are extracted from MODIS data, such as Enhanced Vegetation Index (EVI), four texture characteristics, and Land Surface Water Index (LSWI). Compared with the classification results, we think that SVM and DTs methods each have advantages, and also can achieve high classification accuracy. The overall classification accuracies are 87.68% and 89.13%, Kappa coefficient of 0.86, 0.87, respectively. The result shows is DTs method with slightly higher classification accuracy is more suitable for the land cover/use classification of the study area than SVM.