| The use of satellite SAR to
monitor agricultural land.
It is often impossible to classify land use on the basis of an
individual image due, for example, to poor resolution. However,
classification may be possible based on exploiting the temporal
fluctuations in the returns from different types of scene. This
example discriminates between uncultivated (e.g. woodland or
set-aside) and cultivated agricultural land using optimised joint
segmentation and temporal texture detection on ERS imagery. It
also demonstrates the ability to detect change in land use.
The results shown demonstrate that optimised temporal
fluctuation detection (based on normalised log data) allows one to
classify unchanging and changing land use areas. Not obvious
(since the fixed filter results are not shown) but true: they
demonstrate again that segmentation offers an optimised adaptive
filter to identify the regions of constant RCS, and that the
segmentation results are much better than can be obtained with a
non-adaptive filter.
|
