ABSTRACT
Multitemporal analysis for urban growth usually requires single classification or segmentation of all the stacked images and is limited to provide detailed change information because of low temporal resolution. Therefore, time series imagery applied in differentiating land cover has attracted increased attention from researchers in recent years, because temporal domain has showed its advantages in resolving class confusion between classes with similar spectral characteristics (Bhandari et al. 2012; Schneider 2012). Specifically, Landsat time series have been successfully applied to map dynamics of urban areas because of their long record of continuous measurement at effective spatial resolution and temporal frequency (Gao et al. 2012; Li et al. 2015; Sexton et al. 2013a,b). However, these methods focused on spectral differences or temporal consistency after classification. Little attention was paid to temporal data mining method to differentiate urban areas from other land cover using dense time series Landsat images.
