I am hoping to extract some simple vegetation features (sagebrush) from 1m 3-band (RGB) NAIP imagery. Unfortunately, there is no near-infrared band available for this dataset and I need to use this particular imagery for a time-series analysis, so I am stuck with the 3-bands. If this were 4-band imagery, I would consider adding NDVI and EVI vegetation indices as ancillary data for the classification. I do plan on incorporating texture into the classification.

What additional band indices or useful information from widely available data (e.g. NED, landform) can I incorporate into the classification to increase the accuracy? I am flexible in what classification approach that I take.

  • 1
    I'm not very familiar with sagebrush but is there some kind restricted environment that it grows in you could incorporate measures of relative landscape position derived from DEMs, e.g. elevation above stream to map the low, or high, or mid-slope positions that it may grow in. Or perhaps if it prefers to grow in slightly wetter soils, you could include a DEM derived curvature raster in the classification to find convergent slopes. – WhiteboxDev Sep 5 '14 at 22:04
  • All good ideas @WhiteboxDev. Sagebrush grows in a wide range of arid regions in the western USA. – Aaron Sep 5 '14 at 22:13
  • 1
    Good luck with the classification. You always ask such interesting questions! As a further note, by incorporating measures of relative landscape position, you can tap into some of the biophysical processes the control sagebush's spatial distribution. – WhiteboxDev Sep 5 '14 at 22:16
  • 1
    As you say, you may also generate texture patterns as additional synthetic channels, even with different moving window sizes. – markusN Sep 6 '14 at 0:04
  • Check this: [How to extract surface from air photo avoiding trees and buildings?][1] [1]: gis.stackexchange.com/questions/109413/… – Pau Sep 6 '14 at 17:00

I did this type of thing for a college project some years back using 25cm aerial photography. It is a difficult thing to accomplish. I ran a number of texture analysis on the imagery and added the bands to the RGB imagery to have more information during the classification process. While it is not a substitute for the NIR band, it did provide some additional information that increased the classification accuracy.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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