Characterization of micro-terrain features has been explored to detect gully objects in the terrain. An adaptation to the morphological reconstruction operator is implemented to detect gullies instead of buildings or other man-made structures.
This operator can be configured to different gully depths and widths. The algorithm is based on successive geodesic dilations applied on a moving kernel. The geodesic dilation uses a mask (shifted copy of the original terrain) to generate a reconstructed surface,
which ultimately can be subtracted from the original terrain to produce off-terrain elements or gully zones. Thus, the algorithm uses as inputs the original DEM, and a predefined height (mask shift) and width (kernel size) in meters, to
customize the minimum detectable gully by the operator.