23 Nov 2015 07:57 AM |
|
<p>I am having trouble radiometrically calibrating WorldView-2 multispectral imagery. In the past, I was able to do so using ENVI's Radiometric Calibration tool to produce nice results with appropriate spectral profiles. Now, my results for different images acquired at various times of year in different study areas are all unusable. Even using the exact same approach in ENVI with the same image that produced good results before now gives me unusable results. I have tried ENVI, IDL and even Python calibration techniques (all black-box techniques, not of my own code), and none of them produce usable calibration results. I don't seem to be able to upload images or screenshots of the good and bad profiles, but could someone please offer a list of troubleshooting steps I should try? To reiterate, I used ENVI's Radiometric Calibration technique using the exact same steps on the exact same image that produced good results in the past, and now have unusable results.</p>
<p>Thanks in advance for any help.</p>
|
|
|
|
Deleted User Basic Member
Posts:228  
25 Nov 2015 11:06 AM |
|
I wonder whether perhaps you have chosen a different calibration setting than you used before? ENVI's Radiometric Calibration tool can calibrate to either radiance or top-of-atmosphere reflectance. The results would be very different if you chose different settings. Perhaps before you used a different setting here? Also, perhaps you changed the Scale Factor parameter, or the output data type? Or perhaps before you chose to Apply FLAASH Settings, which creates output in the radiance units needed as input to ENVI's FLAASH atmospheric correction tool.
Can you describe in more details how the results differ from the previous results? For example, are the pixel values in a different range than they were in the previous calibration? Are the spectra shaped very differently? Also, were you using a different version of ENVI before, or is this the same version of ENVI?
For what it's worth, I just calibrated a WorldView-2 dataset to radiance using the Radiometric Calibration tool in ENVI 5.3, and output using FLAASH settings. The results look like I would expect.
- Peg
|
|
|
|
Deleted User New Member
Posts:  
07 Dec 2015 11:42 AM |
|
Thanks for your reply, Peg. Sorry for the delay in responding. I double-checked the settings and all are the same as before. Moreover, I've used correction algorithms in ENVI, IDL and Python, some of which are independent of the others. I also updated ENVI to 5.3 and tried again, but found the same poor results.
There appear to be two primary differences in the good and bad spectral profiles: Radiance values are higher in the bad profiles by 1 order of magnitude (e.g. ~50 in bad vs ~5 in good); and the red band (band 5) appears to be normalized relative to the other bands as it varies minimally from pixel to pixel while other band values fluctuate more realistically.
Given the consistent results between independent programs, the source of error may be in the calibration metadata, but I don't see how that could have happened to some files and not others.
|
|
|
|
Deleted User Basic Member
Posts:228  
08 Dec 2015 01:23 PM |
|
I agree that if you are seeing similar bad results with different, independent tools, the problem sounds like it must be more related to the data than the tools. The way you describe your red band sound unusual, and problematic. You might want to contact whoever provided the data to you (or whoever the original data provider is), to see if they have any ideas about what might have gone wrong.
- Peg
|
|
|
|
Deleted User New Member
Posts:  
09 Dec 2015 08:25 AM |
|
Okay, will do.
|
|
|
|