I am an expert on my field.

I work with remote sensing. Mostly UAVs. I collect datasets and extract features from it. One of the features is 'green canopy cover' (https://acsess.onlinelibrary.wiley.com/doi/10.2134/agronj15.0150).

The formula is here:

pixel based clasification

This is pixel based clasification. For the last 3 years, I have used DJI P4 RTK for collecting the datasets (20 MP camera). This year, I started collected with the Astro Freefly and the Sony A7RIV sensor (61 MP camera). I have noticed the images from the Sony sensor are a bit darker in comparison with the ones from DJI. Canopy cover is calculated from the generated orthomosaic. I always follow the same procedure on Agisoft Metashape to generate it.

Both are mosaics, from DJI and from Astro, are 8-bit integer.

See here how the same green pixels are left undetected. This happens on the mosaics from Astro and not from DJI.


green canopy cover

How calculating canopy cover differs from the sensor utilized?

  • I can't find proper stats on the Sony cam but this could be a difference in band wavelength. Basically one sensor collects a different wavelength area per color band than the other.
    – user90323
    Commented Mar 31 at 16:56

1 Answer 1


I have contacted paper's author. He has expresed the following: "The resolution is not adequate to accurately measure the green canopy cover. The size of one pixel needs to be smaller than the smallest dimension of the leaves you are trying to measure. Still, with the Canopeo algorithm you can adjust the thresholds to try to get more accurate results. The default values of 0.95 are not always the optimal values."

This is not a sensors problem as I suspected in the first place. I have confirmed this by measuring green canopy cover a week later: green canopy cover binary classification

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.