[GIS] Process of Radiometric Calibration

multi-bandremote sensing

I am trying to understand the process of Radiometric Calibration of multi-spectral images acquired using Parrot Sequoia sensor with firmware v1.2.1.

Given a radiometric calibration target and the Albedo values for all bands of interest (Red, NIR, Red-Edge), how to apply this information to get a radio-metrically calibrated image.
What is the formula to apply radiometric calibration?

Best Answer

With more than one calibration target, you can estimate the relationship between your measurement and the albedo/reflectance with a linear regression.

If you have only one calibration target, you need some more assumptions and I recommend you to use simple equations to avoid overfitting. Those are applied for each band individually.

  • With a clear sky, the safest is to try to remove the bias of your reflectance measurement (ideally using a dark body as a calibration target, which is then used to remove noise). To do so, add the difference between 1) the average pixel values on the observed calibration target and 2) the theoretical values of your target to all your observed pixels. This calibration method takes the atmospheric effect into account.
  • if you have clouds and/or aerosol issues, then you can multiply all your observed pixels with the ratio between the average pixel irradiance values on the observed calibration target and 2) the theoretical values of your target. Be carefull that this will yield incositent results on partly cloudy images.

This assumes that your software provides you with estimated reflectance/ irradiance values based on internal camera calibration parameters and (for reflectance) using the sun position, sun-earth distance and sun spectrum.

Related Question