New approach for spectral change detection assessment using multistrip airborne hyperspectral data

Change detection of imaging spectroscopy data is widely used in many applications. Among them, environmental monitoring is of great importance. In this paper, we introduce a new automated method, termed spectral overlapping threshold (SOT), to derive a threshold to distinguish between 'change' and 'no change' areas. The method exploits the overlapping regions in multi-strip mosaic images, which are regarded as 'no change' areas because they are acquired only a few minutes apart. The method consists of two steps. First, similarity measures are applied to the overlapping areas. Then, the histogram of the similarity values are computed and the thresholds for each land use land cover (LULC) category are determined. The method is independent of the underlying SM used to detect changes, and is demonstrated here for the spectral angle measure (SAM), spectral information divergence (SID), Euclidean distance (ED) and spectral correlation measure (SCM). This process is demonstrated for a mosaic of HyMap sensor data acquired in 2009 and 2010 over Sokolov mining area, Czech Republic.