Best settings for 3D buildings and terrain

Hey everyone,

I am trying to figure the best settings (ODM settings and drone flight path) to model an old roman ruin and the terrain around it. I am more interested in a good cloud points and textured model, the orthophoto seems already very good.

So far, here are my results:
_1st attempt: flight path was: 1st grid at 40m AGL, camera gimbal pitch at 85° below the horizon, 70% front overlap, 65% side overlap. 2nd grid at 50m AGL, 20° offset direction from 1st grid, camera at 85°, 70% and 65% overlaps. 1 orbit at 30m AGL with a 45° gimbal pitch camera.
As you can see here:

the terrain is very well rendered but there is some holes in the 3D building, and if we zoom in there is a lot of errors near the windows and gates, the building is not very detailed.

_2nd attempt: same pictures as in the 1st attempt + pictures of the facades taken at 14m at 0° gimbal pitch. Here is the result:

This time the 3D model of the building is very good (only a few minors errors) BUT now I have gaps in the terrain.

On both projects, WebODM settings are:

auto-boundary: true, boundary: {…}, dem-resolution: 2.0, dsm: true, dtm: true, feature-quality: ultra, mesh-size: 300000, orthophoto-resolution: 2.0, pc-geometric: true, pc-quality: high, use-3dmesh: true

Am I doing something wrong with ODM parameters or the drone flight path ?
How can i get the terrain of the 1st attempt and the building of the 2nd attempt at the same time ?

Here is the survey data for comparison:

I am using webodm docker version on a Linux server.

Cheers,

David

3 Likes

Datux - I’m a noob, so excuse my lack of knowledge.

  1. What was the resolution of your drone camera - 2K/4K/6K/8K or something else?
  2. How many images did you collect and would it be possible to combine both attempts under the same project?

Here’s my take. If your drone was shooting in low resolution, then it was not able to capture as much information for the render. If you were able to combine both attempts, then information missed in one might be captured by the other. I just shot my neighborhood and collected about 20 images…they rendered, but not nearly as precise as I was looking for. I’m going to shoot it again, come lower, fly slower and collect more shots.
~
EM

2 Likes

Hey! Welcome!

It is typically recommended that overlap/sidelap for detailed 3D Models be 85% or greater, you’re a bit lower than that with your stated flightplans.

You might also benefit from pushing --pc-quality ultra and raising --min-num-features 25% or greater.

Thanks for the replies,

EM, I am a noob too, so no problem: My drone camera has a 4K resolution, pictures are at 4000x3000px. 84 pictures were taken for the terrain, so it’s an average GSD of 1.4cm/px (not bad I will say) plus 54 pictures for the facades for an estimated GSD at 0.44cm/px.
Combining both projects ? I don’t know, theirs absolute position are a little off if I overlay them in QGIS. And I do not know how to do it for the textured objects, so that will be a last resort option.

Saijin_Naib I was shooting a double grid at 70% for an equivalent of a single grid at 83% as recommended here: https://docs.opendronemap.org/tutorials/#calibrating-the-camera . For the min-num-features option it require an integer (default at 10000) not a percentage, so what number should I try ?

If I had to rephrase it, my main question is: Why gaps appears in the terrain after adding facade pictures ?

2 Likes

Thanks for the info. You’ve given me some ideas that I wasn’t aware of and now I’m going to test them out.

  1. Can I combine two flight captures, ortho and facade?
  2. What is the optimal image overlap? Too close will create distortion and too far will create gaps.
  3. Is there an optimal resolution? 4K is 12MP, 6K is 20MP, and 8K is 48MP. Does increased resolution improve results?

I’m piloting an Autel Evo II 8k so these will be questions worth answering to me.

EM

1 Like
  1. Yes, provided there is good overlap between the two. Change of gimbal angle helps a lot.

  2. You can’t go too high, really.

  3. The more the better, technically. Your hardware limits the benefits. Also, many newer sensors are quad-bayer and their real resolved resolution is roughly 1/4 the stated max, so pushing it higher doesn’t really do much benefit.

1 Like

Thanks for that. When (if) the weather clears, I’ll work on some new captures. My hardware limitations show that I can’t process 291 images with only 16Gb of ram. :confused: So I’ll adjust.

1 Like

this page can help , someone post it to help newbes as my case, https://support.pix4d.com/hc/en-us/articles/202557459-Step-1-Before-Starting-a-Project-1-Designing-the-Image-Acquisition-Plan-a-Selecting-the-Image-Acquisition-Plan-Type#label4. I on orbit mode take 36 pictures minimun ,1 every /10º as advice the tutorial. Sorry is for Datux

1 Like

imagen


As you can see to do front shots done gaps on x axis ,i presume that there are less x key points .

2 Likes

This was helpful. I went a little slower about 197 ft and took 139 photos and came out with this 3D image.


8 Likes

Very, very well done!

How can this error be explained?
It makes the orthophoto of the facades and as you can see the positions of the camera are offset in half of the photos…it is a low cost fimi x8 2020 drone



Capturafallo3d2

1 Like

Thanks ricardoarce for the link, very informative. According to this documentation, my pictures of the facades at 0° gimbal pitch was probably not a good idea. I will try a new flight path when ever possible.

2 Likes

Probably very poor GPS accuracy and consistency.

1 Like

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