Supervised farm classification from remote sensing images based on kernel adatron algorithm

The main focus of this paper is to propose a new supervised farm classification method from remotely sensed Landsat7 ETM images and based on the kernel-adatron (KA) algorithm. This algorithm produces the separation of two farm classes by an optimal decision boundary defined by a linear separating hyperplane in a general feature space. Nonlinearities are handled by mapping the input data into a multidimensional feature space induced by a kernel function. The experimental results suggest that effective farm classification based on spectral characteristic recorded in a satellite image is possible; and reveals that repeatable relations between biophysical and spectral features can be derived from abstractions difficult to observe as farms.