Mapping Vegetation at the Species and Individual Tree Level

Classifying high resolution imagery by multispectral signature is more difficult than with standard satellite data because of the increase in variability created by the greater spatial detail. We use new object-based classification routines that reduce this variability by segmentation while retaining the useful information necessary for detailed classification and mapping at the species level. Also, the same delineation routines that reduce variability by creating homogeneous polygons can be used to automatically measure the individual tree crown diameters in stands of forest.

Example of image classification using object oriented feature extraction in Hawaii Volcanoes National Park, Hawaii