| 19 Jan 2007 07:09 PM |
|
|
I and a number of colleagues have become interested in repeat photography for monitoring lichen growth and cover. Photos could be close-ups of lichen covered rocks, or slightly more distant photos of trees draped with lichens. I've been playing a little with classifying some images, but so far it seems that shadow variation causes a lot of error.
Does anyone have experience with similar situations that they would like to share? Any citations of papers I should read?
Suggestions on transformations, texture measures, and classification methods would be most welcome!
Thanks,
-Eric
|
|
|
|
Deleted User
New Member
Posts:  
| 19 Jan 2007 07:09 PM |
|
|
You could a few different approaches:
- Band ratios, to remove the effects of topography/shadows. Then apply any classification technique you like (Max Likelihood, SVM, etc.).
- Spectral Angle Mapper classification, which is insensitive to topography/shadows
The precision/accuracy of the classification is heavily dependent on how many bands of data you have to work with. If the images are three-band RGB, you might still run into problems since the dimensionality is limited.
|
|
|
|
Deleted User
New Member
Posts:
| 19 Jan 2007 07:09 PM |
|
|
Hi Eric,
Did you make any progress on this? I have a similar situation with repeat photos of a ground plot. I, too, experimented with image classification, with moderate success. I'd love to hear if you got any helpful responses or suggestions.
Thanks,
Theresa
|
|
|
|