Drone/Image Height for Ortho

Hi

I took some photos of a Cauliflower field from around 2m high. I uploaded the images onto WebODM and an orthomosaic image was created. The ortho however doesn’t have near the level of detail I require to estimate the number of vegetables I have growing.

Is there a height limiter? Is there a way to change this?

I have shared 2 dropbox links, one shows the output I get from WebODM and the other the photos I collected using my drone. You will clearly see the difference.

I’d really appreciate any help and guidance.

I use WebODM and DockerDesktop on a Mac running Monterey Version 12.3.1. Please let me know if any further information is required.

Dropbox link to WebODM Ortho:

Link to Image FIles

Many Thanks
Bass

2 Likes

2m is far too low. I’d suggest 5mm GSD is plenty to count even small cauliflower plants, so depending on what drone you have 15 or 20m would give a better result

1 Like

to add just a bit to what Gordon wrote. I would suggest changing aspect ratio to 3:2 with air 2s instead of 16:9, with 16:9 you are automatically cropping the image.
with that camera from 20 meters your GSD would be around 5.5 MM and would need to space images out about 7.5 meters for a side overlap of 75%.

hope this helps.

2 Likes

Guys thanks so much for your feedback, super useful. We flew low yesterday becuase aside from stand count we are trying to find out how many of our plants are unproductive and we thought that flying higher wouldn’t be sufficient.

So I understand, what prevents the ortho stitching from working when flying low? Why am I getting such a pixelated image?

3 Likes

This is the textured point cloud, which turned out better than the ortho. Ortho’s do not generally reproduce well with tiny GSDs.

Camera distribution isn’t ideal, you really want consistent forward and sideways overlap

56images 00:50:54
|Processing Node:|node-odm-1 (auto)|
|Options:|auto-boundary: true, feature-quality: ultra, gps-accuracy: 6, mesh-size: 250000, orthophoto-resolution: 0.1, pc-quality: high, use-3dmesh: true|
|Average GSD:|0.07 cm|
|Area:|89.39 m²|
|Reconstructed Points:|6,002,016|

1 Like

Thanks Gordon, I have tried to re-do the same locally but am getting ‘not enough memory’ with a sample of 5 photos.

I guess my computer isn’t up to it. I will either try to play around with the settings and reduce the burden or try to fly higher on my next visit.

Thanks again for your help!

2 Likes


I am able to process it.
but don’t know why the result is so small.

use:
dem-resolution: 2.0, fast-orthophoto: true, feature-quality: ultra, mesh-size: 300000, min-num-features: 20000, no-gpu: true, orthophoto-resolution: 2.0, pc-geometric: true, pc-quality: ultra, resize-to: -1, rerun-from: dataset

2 Likes

report (3).pdf (2.4 MB)

I would recommend fly much more images if you need this solution.

if it is DJI Air 2S, you should able to use some app to auto flight with 85%+85% overlap.

3 Likes

Maybe because you have set resolution to 2cm, when it is actually <1mm, but I find orthos don’t reproduce well when you have tiny effective GSDs.

3 Likes

thank you. I run it gain. and get 725MB ortho :smiley:

2 Likes

great
Screenshot 2022-07-14 173039

2 Likes

I’ve got plenty of orthos that are many GB in size :wink:
How many cauliflowers did you count? :rofl:

2 Likes

I don’t like to count it manually. Do you have any other option?

2 Likes

If it was something I needed to do, I’d draw a grid over the field, say 1m X 1m, and count the plants in a few representative squares, take the average of them, then multiply by the total number of squares, adjusting for any part-filled squares around the edges.

3 Likes

For cauliflowers, looks like they are not mean spread.

I just do mine. The difficult to count the trees by size or by canopy.

3 Likes

If you know how many you planted, it would be easier to count the missing ones.

2 Likes

https://www.opendronemap.org/fieldimager/

4 Likes

Our ecosystem never fails :smiley:

3 Likes

This is very difficult for me. I don’t have any experience in all this.

I have done these.

  1. Install Download R-4.2.1 for Windows
    https://cloud.r-project.org/
  2. Install Rstudio
  3. Rtools
  4. Devtools

and downloading this now. Docker Hub.

In the https://www.opendronemap.org/fieldimager/
after I try all, it looks like install everything already.
I am using windows 10 with webodm docker version.

In the link say: 1. First steps
Taking the first step:

library(FIELDimageR)
library(raster)

I am able to run it now. But it still looks very complex for me. :exploding_head:

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