[GRASSLIST:7279] problems with r.texture

andrewhaywood at telstra.com andrewhaywood at telstra.com
Wed Jun 22 20:56:07 EDT 2005

Dear all,

I am new to GRASS/REMOTE SENSING/and Linux but I am enjoying the challenge of getting it to work. I am currently having problems using r.texture in the context of a forest health remote sensing project.

I am running GRASS5.4 on a Linux Fedora 3 system.

I have sucessfully imported a 4 band envi format DMSI image. THis image has 4 bands at 2m resolution over a 30 000 ha scene. Approximately 74 million pixels.

I have 200 forest sites where 20m radius plots were measured and forest health was scored on a scale of 1 to 4. I would like to model this health score against some image textural attributes. 

I did not want to run r.texture over the whole scene. So I buffered the 200 "sites" by 100 metres and created a vector file. I then converted this vector file to a raster. I then used this raster to mask the original DMSI image data (its 4 bands). 

My problem is I although I am able to run r.neighbor over a window size of 19*19 accrossed the masked image image, whenever I run r.texture i get a segmentation fault. I suspect I may be running out of memory, 

*Firstly how can I tell if I am running out of memory
*should I break the scene up to smaller areas. Basically I have 200 small areas of interest. If this is the case how can I script this easily.
*is it possible to turn on some type of debugging to see where its falling over

Note if I run i.texture it runs fine. But I really want texture attributes that are at a similar scale to my health plots.

Any help or guidence on this would be much appreciated



More information about the grass-user mailing list