X
PrevPrev Go to previous topic
NextNext Go to next topic
Last Post 22 Apr 2009 08:06 AM by  anon
maximum likelihood classification threshold problem
 1 Replies
Sort:
You are not authorized to post a reply.
Author Messages

anon



New Member


Posts:
New Member


--
22 Apr 2009 08:06 AM
    Hello, I am a new user to ENVI and am having difficulty using the probability threshold function within the maximum likelihood classification module.  No matter what threshold I set (I have tried values from .0001 to .9) it doesn't seem to be able to classify any pixels (in other words, the return is an entire image that is "unclassified" except for masked out pixels.)  I am using a mask band, but I don't see how this would make any difference.  I am also working with radiometrically and atmospherically corrected imagery, so pixel values are on a scale from 0-6000 or so (reflectance in percentage, multiplied by 10000), which are the values returned by the FLAASH module.  Should I change the units of the image before processing?  Or perhaps I simply have the wrong units for the threshold values?  Note, other supervised classification techniques with "max distance error" functions (such as Mahalanobis classification) have been no problem.  Thank you for any replies,

    MariM



    Veteran Member


    Posts:2396
    Veteran Member


    --
    22 Apr 2009 01:27 PM
    Perhaps you need to set an appropriate scaling factor as described in this tech tip: http://www.ittvis.com/ser...echtip.asp?ttid=4059
    You are not authorized to post a reply.