Automatic discrimination of farmland types using IKONOS imagery

The aim of this paper is to introduce an approach for the discrimination between the farmland types cropland and grassland from single satellite images by estimating the main direction of cultivation in cropland. The approach uses structural features caused by the cultivation, in particular straight lines caused by agriculture machines. The core of the approach is the transformation of an edge image into Hough space, and a following interpretation of the results to determine the main direction. The new approach is presented in detail and is illustrated with the help of examples. The examples and the evaluation demonstrate the potential and the limits of this approach.