| 13 Feb 2016 12:14 AM |
|
|
I have ROIs from a high resolution imagery and i want it to compare with Tasseled cap/NDVI/NDBI. My question if we analyse the pixel values (lets say, 0.187) of Tasseled cap/NDVI/NDBI against those ROIs, and then classify these Tasseled cap/NDVI/NDBI using "Raster color slice in ENVI", then generated confusion matrix/kappa coefficient in this way is consider as "Correct" or not?
This process is acceptable scientifically or not?
|
|
|
|
Deleted User
Basic Member
Posts:228  
| 03 Mar 2016 12:55 PM |
|
|
Hi Atiqa,
I wouldn't want to be considered the final word on what is "scientific" or "acceptable", but your approach does seem reasonable to me.
- Peg
|
|
|
|
Deleted User
New Member
Posts: 
| 05 Mar 2016 01:15 AM |
|
|
Although, it seems good to me also... But i have no proof for that...
Thats the main issue.
|
|
|
|
Deleted User
Basic Member
Posts:228
| 11 Mar 2016 10:49 AM |
|
|
I think perhaps the problem is that there is no "proof" that a method is scientifically valid, other than making sure that it logically measures what you are trying to measure. There are many different ways that you could try to measure the validity of Tasselled Cap results, and this does seem like one that makes sense to me. I would not want to tell you that there is not a better way, though. If you are unsure, then perhaps it would help to consult the scientific literature to find examples of how others have done this?
- Peg
|
|
|
|
Deleted User
New Member
Posts:
| 12 Mar 2016 02:16 AM |
|
|
One of the other ways is through image (Confusion matrix using ground truth image) which worked. I actually performed it. But i dnt have more accurate ground truth image, that's why i want this way too.
|
|
|
|