17 Feb 2014 07:03 PM |
|
I am using FLAASH to apply atmospheric correction to RapidEye imagery. However, the output contains pixels with negative values for reflectance. The percentage of these pixels in different RapidEye bands are: B1=89%, B2=30%, B3=21%, B4=4% and B5=0.5. As you can see, a large number of pixels are incorrect except for the fifth band. Any ideas why this happens and how to fix this issue. Has this something to do with RapidEye images? I can provide a more detailed of the steps I took if need be.
Thanks for helping out.
|
|
|
|
Deleted User New Member
Posts:  
18 Feb 2014 12:32 AM |
|
Scale factor value?
|
|
|
|
Deleted User New Member
Posts:  
18 Feb 2014 04:32 PM |
|
Two scale factor values were used. The first one is specific to RapidEye imagery. DN values of level 3A RapidEye pixels were converted to radiance W/(m2 • sr • μm) by multiplying the pixels by 0.01. Then, prior to using FLAASH a scale factor of 10 (multiplied by 0.1) was used to convert the radiance from W/(m2 • sr • μm) to μW/( cm2 • sr• nm). In FLAASH Water Retrieval and Aerosol Retrieval options were set to none because of the lack of these bands in RapidEye. Zenith and Azimuth angles in the metadata was used in advanced setting. Still the negative values I reported were produced.
|
|
|
|
Deleted User New Member
Posts:  
19 Feb 2014 01:44 AM |
|
I processed the scene RapidEye image level 3A using ENVI 5.0 classic sp2. I did not conversion form DN to radiance (only perform BSQ to BIP conversion) and use scale factor for all bands 1000. Zenith (180-|spaceCraftViewAngle|) and Azimuth angles (azimuthAngle) from the metadata was used. The image have a small amount of lakes and clouds with shadows so the negative values can be. The most part of image is covered by vegetation.
The result looks pretty good.
Before FLAASH
Dims: Full Scene (25,000,000 points)
Basic Stats Min Max Mean Stdev
Band 1 3331 20167 4472.558649 500.494362
Band 2 1911 20897 3307.275833 607.266273
Band 3 886 19936 2352.556299 780.088903
Band 4 699 16698 3106.273342 723.337874
Band 5 480 14958 3931.542122 748.104430
After FLAASH
Dims: Full Scene (25,000,000 points)
Basic Stats Min Max Mean Stdev
Band 1 182 7182 656.776141 208.088265
Band 2 144 7693 698.937411 241.464225
Band 3 47 7737 638.937628 314.918884
Band 4 47 7259 1132.113448 326.054213
Band 5 91 7799 1928.183526 398.283757
So i dont know whats wrong with your data.
|
|
|
|
Deleted User New Member
Posts:  
20 Feb 2014 01:38 AM |
|
Thanks Alex for giving it a shot. I followed your processing steps and things got better. It seems that less processing should be done before applying FLAASH (I didn't subset the image prior to FLAASH and used 1000 as scaling factor in FLAASH). Here are stats for my image. Do you think that these average values are OK? FLAASH output is multiplied by 10000 therefore a typical pixel in my image has a reflectance of 0.02 or almost zero. The scene is from Victorian Alpine region covered with forest and for pixels including vegetation I expect a larger reflectance number specially for band 2.
Before FLAASH
Dims: Full Scene (25,000,000 points)
Basic Stats Min Max Mean Stdev
Band 1 2056 14634 3017.035913 176.708060
Band 2 1168 14185 1915.344401 249.527780
Band 3 406 11327 1101.681892 250.215876
Band 4 333 11019 1494.851358 479.294036
Band 5 148 10208 2892.161012 1187.590239
After FLAASH
Dims: Full Scene (25,000,000 points)
Basic Stats Min Max Mean Stdev
Band 1 -341 6968 216.517715 102.683258
Band 2 -185 6965 224.360649 137.049557
Band 3 -219 5797 163.053931 137.757283
Band 4 -188 6861 577.670941 316.168025
Band 5 -142 7386 1910.852001 888.862493
|
|
|
|