Spatial clustering-based parametric change footprint pattern analysis in Landsat images

Spatial data represent the geomorphological phenomenon occurring on the earth’s surface. Several geological phenomena undergo changes over a period of time due to natural and man-made reasons. A footprint is defined as the spatial extent of any geomorphology, ecology or human activity at a particular instant of time. Change detection includes the study of change in footprint of a particular class. In this work, we proposed a footprint extraction method, footprint extraction using spatial neighbourhood based on spatial neighbourhood property. The proposed footprint extraction method proves to be highly effective than different state-of-the-art methods in terms of silhouette score value. An unsupervised change computation framework has also been proposed. The experiments are performed on five temporal Landsat 5 thematic mapper images of the Delhi region. Spatial polygons have been used to identify the spatial footprint of the predetermined class. Some non-spatial parameters like the size, the area, and percentage of area out of the total area of the targeted class have been used to study the extent of a predetermined class. The temporal change vector has been proposed to find the temporal change pattern in the observed concepts.