|
Image processing
software can use the differing spectral properties of land
cover to ‘classify’ a multispectral satellite
image very rapidly. Land cover maps for large areas, which would
take many man-years to survey from the ground, can be produced
in just a few hours. Classification accuracy can be improved
using a small ground verified sample to train the software to
recognise each cover type of interest. This ‘training data’ gives
a measure of the variability each land cover type, such that
a likelihood of each image pixel belonging to each land cover ‘class’ can
be determined and thus the image pixel can be assigned to the
most probable ‘class’.
Image classification can
be used for a great variety of applications. In addition
to producing
a more understandable map, the digital data readily provides
area statistics and by using images from different dates,
a means to detect small changes in land use. This classification
of the
area around Seattle, USA was used together with elevation
data
to model the transmission of mobile phone signals. The propagation
of microwave signals for communication is greatly effected
by the absorption, scattering and reflection of the signal
from
vegetation and urban structures in addition to topography.
Using such data the signal strength from a transmitter network
can
be modeled before it is actually established.
|
