|
|
Methodology.
|
|
- Individual ERS images are not sufficient to classify
forest or crop type because the resolution is low and
the image is corrupted by speckle noise.
- Joint segmentation derives a common set of
boundaries between homogeneous regions in a complete
sequence of images. It is thus much less affected by
speckle noise.
- The temporal signature of average intensity can then
be used to classify each segment using a Maximum
Likelihood classifier. The time sequence provides an
additional important source of information about the
forest or crop type.
|
|
|
|
|
