reply to Dominique

Hong Zhuang zhuang at zorro.cecer.army.mil
Tue Jul 21 16:50:20 EDT 1992


Dear Dominique Bachelet:

        You have asked the following questions:
----------------------
I have been trying to perform an unsupervised classif of avhrr images of Asia
using i.group, i.cluster and i.maxlik.
When I run i.cluster, it generates a long output file with no comments, no units
to the numbers generated etc.  Does anyone know what they all mean, what....
Also how do you pick the initial values so that ......
----------------------

        As I understand,  the units of these numbers are the same as your
input map layers.  For example, if your subgroup includes four layers:
        elevation (in meter),
        satellite remotely sensed radiance band 1,
        band 2 and
        band 3.
Then after clustering, you will get four numbers for means and standard
deviations for each class.  The first number has unit as elevation (meter) while
the others have units as radiance.

        The initial values for iteration are found by computer its self.  There
is no need for user to pick them.

        If we use state space, we will feel easier to understand the entire
process of classification.  In the above example, the state space is a four
dimensional space.  The four coordinate axises are elevation, radiance band 1,
radiance band 2 and radiance band 3.

Hong C. Zhuang
SASteam, CERL
Champaign, IL 61820
Zhuang at zorro.cecer.army.mil
(217)352-6511 ext.588




More information about the grass-user mailing list