1

I have a single GeoTIFF created from some NGB aerial photography that has been stitched in Agisoft Photoscan to create a single image of an area of heath.

I would like to perform spectral mixture analysis on the GeoTIFF but am lost as to where to start. I have sample photographs for end member selection. Can anyone point me in the right direction please?

Improve this question

1 Answer 1

Reset to default
4
+100

Spectral mixture analysis / sub pixel analysis is designed for hyperspectral data, not a 3-band aerial photograph. However, you can try it and see if the output is useful. A tutorial can be found in this pdf and in this ppt/pdf. You will have to skip a significant number of the steps, as you don't have the same amount of information in your dataset. In essence, you can skip the preprocessing steps and go directly to assigning your endmembers based on the geolocation of your photos. Main issue that you'll most likely run into is that you'll want more classes than your 3-band data can actually provide.

Improve this answer
3
  • Thanks Mikkel. This looks useful. Is there another image analysis technique that might be better for classifying vegitation using RGB/NGB aerial imagery?
    – Mark Williams
    Commented Feb 12, 2015 at 13:01
  • 1
    @Willburt, when limited by spectral information, you can start looking at spatial information. You can do this by looking into object based classification / image segmentation. The entry barrier is rather high for this, and it may require a fair amount of effort to get into. Personally, I use eCognition for this type of work, but Orfeo Toolbox (as part of QGis / OSGeo4W) should also be able to do it.
    – Mikkel Lydholm Rasmussen
    Commented Feb 12, 2015 at 13:18
  • Youre right Mikkel. SMA wasn't very useful however your info guided me towards OBIA. So I'm looking at a combination of spatial and spectral classification now. Thanks :)
    – Mark Williams
    Commented Feb 14, 2015 at 15:14

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.