Detection of storm losses in Alpine forest areas by different methodical approaches using high-resolution satellite data

Based on the detection of storm losses in Swiss alpine forest areas, two different digital classification approaches were compared. In contrast to the pixel based classification we investigated an object-oriented classification procedure. The eCognition software package of Definiens offers this possibility. The comparison was performed for images with different spatial resolution - very high resolution images of IKONOS, and images of SPOT in the sharpened mode. The evaluation of the IKONOS image indicated a significantly higher accuracy for the object-oriented classification approach than for the pixel-based method. The eCognition software handles the high level of detail and the associated high texture better than the pixel-based parallelepiped-algorithm. The quality of the pixel-based approach, which takes into account only the spectral information and some derived data-products is limited for very high resolution images. The classification of the SPOT presented approximately the same results for both methods.