Height difference between dense clouds

[anagellerramser]

Hello,

I would like to clear up some doubts. Just to contextualize again, I'm trying to find the difference in height between my dense clouds (same location, 3-hour difference between flights) in the place where vegetation was removed, but I'm still not sure about the parameters.
How I'm doing it:
1- I bring the two dense clouds to cloud compare (already processed in Metashape Agisoft, theoretically aligned, but in cloud compare I also use the clouds registration tool (the one with the icon red and green)
2- With dense clouds (no rasterized), I go to the 2.5 D Volume tool. I interpolate both empty cells, use 0.07 as step, and use the average height.
I don't know if it's right to interpolate, but if I leave it empty, then to calculate the height of the pixels later in QGIS is a problem, and I need the height as close to reality as possible.
3- After that, I rasterize the tool. I use the same step value, and in active layer I use relative height, but I don't know if I should use this or height grid values, in projection I use average and empty cells I interpolate. So I export as a raster (and export all the options: height, active layers....)

What I need in Cloud Compare is a difference of height as my output. This raster of relative height is the height of dense cloud 1 points - height of dense cloud 2 points? Do you think is precise enough for that I need?

Last Question: When the soybean is still early, it is easy to visualize the hole, so the program can calculate the difference of height, but when the canopy starts to close, the hole is no longer visible and I think the drone can no longer capture this difference in height, do you have any suggestions for a tool that I can use to overcome this problem? because my height at the end of the soybean cycle is lower than at the beginning.

I appreciate your help very much.

[daniel]

I believe your process is good.

It's probably ok to interpolate, especially if the holes are small. This may produce some artifacts here and there though.

Another option would be to interpolate the raster in step 3 instead? This may be more robust (as you only rely on interpolation once?).

Also, in version 2.13 you can use Kriging instead of standard interpolation (in both tools). That may give more robust results.

And regarding your last question, apart from using a lidar sensor on your drone, I don't know how you can reliably extract some points on the ground...