Hi! all,
I wish to classify the landuse from the landsat data of 1989. I created a 531 (rgb) composite and the features were quite noticeable. The relcassification part I tried to use pixel value for creating classes for eg 9320 thru 10367 = 1 semiurban etc.
The problem is that there are other types of areas such as uncultivated fields that also get classed into areas such as semiurban. How to deal with this.
And also is there any method to create polygon out of textures. The urban areas are quite noticeable by their texture or arrangement of pixel values. Is there any floating window formula that can use std dev to create such surfaces, I am not sure.
Hope to get some help!
Thanks,
Ambrish
Twice the fun— Share photos while you chat with Windows Live Messenger.
On Tue, Mar 10, 2009 at 10:59 AM, ambrish dhaka <ambijat@hotmail.com> wrote:
Hi! all,
I wish to classify the landuse from the landsat data of 1989. I created a
531 (rgb) composite and the features were quite noticeable. The
relcassification part I tried to use pixel value for creating classes for eg
9320 thru 10367 = 1 semiurban etc.
The problem is that there are other types of areas such as uncultivated
fields that also get classed into areas such as semiurban. How to deal with
this.
And also is there any method to create polygon out of textures. The urban
areas are quite noticeable by their texture or arrangement of pixel values.
Is there any floating window formula that can use std dev to create such
surfaces, I am not sure.
Two quick suggestions:
- mask out zones which aren't of interest (r.mask)
- use a classifier like i.maxlik or better i.smap to perform a statistical
assignment of pixel values to desired classes.
For an overview, see
http://grass.osgeo.org/wiki/Image_processing#Image_classification
Markus