Preliminary tests of the utility of hyperspectral image data to precision farming

In a test of the utility of hyperspectral image data to precision farming, the authors have collected several inter-dependent data sets at an experimental farm on Maryland's Eastern Shore during the 1999 growing season, focusing on weed identification in corn and soybean crop fields. These data included: hyperspectral image data from an airborne instrument; ground feature location data collected with a differential GPS unit; geo-located radiometer and sun photometer measurements at ground level. In addition a database for the cost of all field inputs, materials and labor, was built to allow evaluation of the economic viability of incorporating hyperspectral data as an additional information source for precision farming. By collecting ground based radiometer and sun photometer measurements coincident with sensor over-flights, they were able to identify and partition sources of variation in the AISA image data so that target signals could be more accurately characterized. They used this to provide atmospheric correction for their efforts to identify areas of weed infestation during the early stages of crop emergence. Later in the growing season they used these methods for identifying different strains of crops in both maturing soybean and corn fields.