Flight plan "framework" discussion

Guys, I know this is not a simple answer topic, can be case-by-case, with a lot of empirical experience and scenarios.

Still I want to get some of your experience and thoughts on that.

I want to have a “framework” to decide how I should configure the flight plan, considering different scenarios.

I am assuming here I will always use GCP, I know in more time / money consuming, but I usually want to have some checkpoints.

The parameters I to set-up on the flight planner are:

  • Flight altitude
  • Side overlap
  • Front overlap
  • Gimbal angle
  • Flight speed
  • Flight direction
  • Crosshatch
  • Follow terrain
  • Perimeter
  • Focus
  • ISO
  • White balance

From some readings—and a little experience—this is my take so far:

Flight altitude

This usually is driven by the precision/accuracy I want in my final product (overall precision is more or less driven by GCD).

The only case where this can change is when measuring stockpiles, where the recommendation is to be 3 - 4 times the stockpile hight.

Side / Front overlap

From my readings is something from 75% - 85%.
This can be less when using Crosshatch and more in situations where you need more details or things are too similar (forest for exemple).

I had no experience or reading on how to differentiate when to use different side vs front overlap

Gimbal angle

For most cases use -85° (nadir + 5° foward).
The 5° usually have very low impact vs nadir and can help mitigate some bowing effect. I got this from @Smathermathe blog

For cases where you want a more accurate object, -75° seems like the “magical number”. Here I have very low experience.

Flight speed

Less is better due to rolling shutter.
Here it will be a balance between time vs quality—not so sure how to really balance that.

Flight direction

Usually I change this based on what will lead to less time flying, but wind speed can play a big role here too.

Crosshatch

As far as I can tell is “similar” to overlap and/or gimbal angle. It will increase the quality of the results.
For object is better to use, so you get more faces. In some forest environment it also make for better results.

Additionally from @Smathermathe blog and the book ODM Missing Guide, using crosshatch with different flight altitudes and gimbal angle can lead to better results and also mitigate bowing effect.

I have not tried these yet, but I am assuming something like:

  • First you fly at 60m with gimbal at nadir in one direction
  • Second you fly at 65m with gimbal at -85° in a perpendicular direction.

There is a more complex approach from @Smathermathe blog that I have never tried and seems very time intensive to do it in the flight planning apps I have uses (Drone-Deploy, Copterus, Drone-Harmony). Maybe using Litchi, can be done, but I am too noobie to use it.

Follow terrain

When possible, use it, it makes GSD more consistent and flight path more secure.

Perimeter

I usually use it, make flight longer, but I get good results—this is more intuitive than actually a fact/experience.

Focus

From reading and some experience, auto-focus always got me the best results.

ISO

Adjust accordingly in the beginning of the flight, but is usually something around 100.

White balance

I use based on the weather.

I really want to get your takes, intuition and experience here.

Cheers
Lupion

4 Likes

I’m investigating flying a “starhatch”, I made that up. It’s flying two more times 60 and 120 degrees from the first flight path. And with camera 10 degrees from nadir.

I hope that will give a very good result.

3 Likes

Frontal Overlap:
I always push this as high as the platform can reliably capture images.

It is very cheap (costing only storage and a bit extra processing time), and can help with ensuring adequate matching and increasing reconstruction accuracy.

If you tune your sidelap right, you can even get away with bumping that lower thanks to higher frontal overlap, saving you time in the air/on-site.

5 Likes

Interesting approach.

Is there a reason for the 10° from nadir instead of 5°, or is just experience?

2 Likes

Oblique images gives ODM more to work with to get a better camera calibration.

3 Likes

@APOS80, @smathermathe has a very clever approach here @Smathermathe blog that similar to what you are trying.

2 Likes

I’ve seen it, it’s great info. I’m just trying to build upon it.

3 Likes

Like you said, I think it is difficult to establish a one size fits all framework for all scenarios. I sort of have 3 frameworks depending what I want get from a dataset:

  1. Orthophoto only
  2. Orthophoto and accurate DEM
  3. Fully textured 3d model

I find that for #1 I don’t need oblique images at all. For #2 I need enough obliques (10-15%) to get a good calibration and maybe a few more in key areas. For #3 I need A LOT more obliques, and captured from many different angles, and that it comes with some sacrifices to the accuracy of my DEM.

I am mostly commonly looking for #2, and I have had a lot of success following the principles in this paper:

I don’t do #3 often enough and haven’t really nailed down my flights practices for this.

4 Likes

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.