X
PrevPrev Go to previous topic
NextNext Go to next topic
Last Post 21 Aug 2017 10:30 AM by  MariM
Applying a mask when using Support Vector Machine
 4 Replies
Sort:
You are not authorized to post a reply.
Author Messages

Edward Kelly



New Member


Posts:6
New Member


--
20 Aug 2017 04:07 AM
    Hi, I am using a mask to mask out areas of NoData, however when I run a Support Vector Machine there is no option to apply a mask and therefore the areas that were previously classified as 'masked pixels' in other classifications (e.g Spectral Angle Mapper) are now classified as a land cover. Is there a way to apply a mask when running a SVM?

    Thanks,

    MariM



    Veteran Member


    Posts:2396
    Veteran Member


    --
    21 Aug 2017 08:40 AM
    ENVI's SVM tool (which is from ENVI Classic) does not support masks or No Data values. However, the new SVM Classifier from the Classification framework does but it requires you use the API:

    http://www.harrisgeospati...mclassifiertask.html
    http://www.harrisgeospati...examplesvmtasks.html

    With this routine, you can pass a MaskRaster:
    http://www.harrisgeospati.../envimaskraster.html

    Matt Spencer



    New Member


    Posts:14
    New Member


    --
    21 Aug 2017 09:49 AM
    Is this only available in ENVI 5.4?

    Matt Spencer



    New Member


    Posts:14
    New Member


    --
    21 Aug 2017 10:00 AM
    Could i not just spatially subset the study area instead of using a mask?

    MariM



    Veteran Member


    Posts:2396
    Veteran Member


    --
    21 Aug 2017 10:30 AM
    Yes, the classification framework API was added in ENVI 5.4. If your mask only covers background area and not pixels within and throughout the image, then yes, you can simply subset out the background that you do not want included. It will still need to be a rectangular area. If you subset via an ROI or shapefile, you can still have 'background' for areas outside of the ROI/shapefile that will not be masked.
    You are not authorized to post a reply.