Clustering of Time Series for Land Change Detection: A Data Mining Approach

A significant field of study in the recent age is the detection of changes in land cover and land use (LCLU), because any conversion of land cover its impacts on the local or the global environment. LCLU change detection is an important topic of research for the earth scientist using satellite data. The enormous amount of earth science data motivates the data mining community to explore for a reliable, effective, and affordable solution to the earth science domain problem (i.e., land cover change detection problem). In this study, segmentation-based techniques for detecting changes in land cover are illustrated. The proposed algorithms employ a two-step process to detect abrupt changes in the enhanced vegetation index (EVI) time series. The first step reduces the dimension of time series by using the fast fourier transform (FFT) on each annual segment and the second step applies the clustering technique to the reduced dimensions of the annual segments. The quantitative analytic approaches have been used to examine the performances of the proposed methods.