Multispectral Data Processing Q and A
Description of the issues
The panel number I am using is RP04-1824214-SC. I do see that the values on the panel 51.0, 51.0, 51.0, 50.7, and 50.9 % so 0.51, 0.51, 0.51, 0.507, 0.509 when converted to decimal values.
However, in my Pix4D, when I import the calibration image, it automatically generates values that are slightly different ( I did not enter the values manually, they are there when I select the calibration image):
0.50825, 0.5074, 0.50632, 0.5046, and 0.50583.
Please see the attached as an example for the Blue band.
Question 1
Which group of values should I use?
Answer
The short answer is that you can choose whether to rely on Pix4D's values or ours; It should not matter which you choose. As long as you stick with one "source of truth" (i.e. don't mix and match) then you'll be perfectly fine. The reflectance calibration will be accurate either way!
The long answer is that we interpolate the reflectance curve slightly differently. Pix4D uses a linear interpolation formula between the center wavelengths of our bands (750nm-850nm), while we do a least squares regression along a wider range (400nm-1000nm) to smooth out the curve. The difference is very small (less than 2% error), but that is why you will see some slight differences.
The measurement from the radiometric target has an error of 5%, so if the difference in the reflectance factor is close to 1-2%, it should be okay. For your case, the values only differ by less than 1% and even for other users, it is never more than that.
Question 2
Are the numbers generated in the Pix4D read from the QR code and then generate the values?
Answer
Yes. However, the data available for the QR code is very limited, so MicaSense uses linear interpolation and just encode the two ends of a line since the panels have a very flat response. If we were to add more data to the QR code, the code becomes too dense to read reliably from the normal panel image distances.
These auto-read values are meant to speed up workflows in most cases, but you can always enter the more precise values before processing if you need that extra precision.
Question 3
When doing the radiometric calibration, sometimes the panel values are not recognized automatically. See this example.
The blue band is recognized as “0” when it’s loaded (figure a). Even I draw a blue color area, it won’t recognize the blue band reflectance factor (figure b). So I manually input 0.51 (figure c). Is that OK?
Answer
Yes, it is fine to do it this way. The blue color region can still generate a good statistics of the pixels as long as sufficient pixels are included.
Reference:
Xu, Y.; Shrestha, V.; Piasecki, C.; Wolfe, B.; Hamilton, L.; Millwood, R.J.; Mazarei, M.; Stewart, C.N. Sustainability Trait Modeling of Field-Grown Switchgrass (Panicum virgatum) Using UAV-Based Imagery. Plants 2021, 10, 2726. https://doi.org/10.3390/plants10122726.
https://www.mdpi.com/2223-7747/10/12/2726