No data values showing up in a supervised classification

1127
3
05-31-2012 10:57 AM
grahamlangley-edwards
New Contributor II
Hi, I have been doing supervised classifications for several Landsat images.  For the last image I classified as usual, but when I ran the classification two of the categories came up as no data.  I have included a screen shot of a the classification with no data.  I have no idea what went wrong with this classification as I followed the same work flow for all of the previous classifications, which ran fine.  I am using ArcGIS 10, and Windows 7 Professional.  If anyone has any ideas I would really appreciate the help.

Thanks,

Graham
0 Kudos
3 Replies
JeffreySwain
Esri Regular Contributor
What could be happening is that the pixels that are appearing as nodata are being rejected by the reject fraction in the classification tool.  Per the help, reject fraction controls the "portion of cells that will remain unclassified due to the lowest possibility of correct assignments."  I would consider whether the training samples you are using would be applicable to those areas and then consider performing the classification again.  If you are applying a signature generated from another set of files, the pixel values could be different.
0 Kudos
grahamlangley-edwards
New Contributor II
Hi Jeff, thanks for the reply.  I'm not using a signature from another file, just performing a normal Interactive Classification for each image. I don't have that much experience with ArcGIS, but I understood that the reject fraction was involved only when performing a Maximum Likelihood Classification.  Is it also involved when running the Interactive Supervised Classification?   If it is, I don't see why it would have happened just for this layer.  Anyway, redoing the classification would take a very long time, is there no other way to rectify this problem?

Thanks,

Graham
0 Kudos
grahamlangley-edwards
New Contributor II
Ok, I got one of the two categories to show up by splitting the category, adding some extra training samples and then running the classification again.  But this same technique did not work for the other no data category.
0 Kudos