TIME SERIES ANALYSIS OF L-BAND SAR FOR AGRICULTURAL LANDCOVER CLASSIFICATION

In this paper, an agricultural land cover classification algorithm is presented that uses L-band SAR observations over time to create a crop/non-crop classification. A statistical measurement known as the coefficient of variation (CV) is introduced, which as a measurement of change is able to differentiate between changing agricultural fields and relatively static non-crop areas. Images from the airborne AgriSAR 2006 campaign in northern Germany and from the ALOS PALSAR satellite over Minnesota, USA are used to demonstrate the algorithm. The small region covered by the AgriSAR campaign is used to highlight the differences in behavior of L- and C-band, and in this case the better results of L-band data when using the CV algorithm. Results in both regions show promise for future use of the CV algorithm.