Variables that influence the quality of the orthomosaic

==========
Hello, I would like to know which variables influence the final quality of the orthomosaic? I’m doing my master’s project on drones and I’d like to know opinions about those variables. Thank you

Hi Javicole4,

I’m not an expert but in my experience those variables are,
orthophoto-resolution
texturing-nadir-weight
depthmap-resolution
mesh-octree-depth and
mesh-size

Regards

What does ‘quality’ mean to you?

@israelbar has a great list of things that impact quality - but without a target to aim for in terms of what quality means, it’s hard to provide any meaningful advice…

1 Like

Helllo adamsteer, by quality in the orthophoto i meant that your orthophoto is 100% complete, has good resolution and has no gaps. I was thinking that the quality of the orthophoto is affected by the flight height, overlap percentage and the resources of the computer. It’s because i’m testing which flights heights and overlap percentages are better for a specific tree plantantion. So i’m looking for the combination of flight heights and overlap percentages that allow me to obtain the best orthophoto.

Thank you very much israelbar.

Regards

This can oftentimes be entirely a function of the flight plan.

What is your camera? What is your GSD? What is your overlap/sidelap? Do you fly cross-grid or at an offset between plans? What is your overfly?

We need to start there and ensure those are appropriate before you start tweaking parameters.

Hi, @Saijin_Naib I tested three different heights ( 80, 90, and 120 m) using two different overlaps (80x80 and 90x90) because I’m looking for a combination of flight heights and overlap percentages that allow me to obtain the best orthophoto for a specific tree plantation. We use a Phantom 4 pro and we flight cross-grid. The GSD is 2,22; 2,49 and 3,3 for the 80, 90, and 120 m height respectively. I hope this information helps you more. Regards. Captura

1 Like

Please, describe the area you are mapping. Is it urban, flat, steep, grassy, etc…???

The area is a tree plantation (5,814 ha), is flat and it´s not very grassy.

What is the height of the trees?

Between 12-14 meters

javicole 4,

I think a 90 to 120 m elevation flight could be right.
Front overlap 80 to 85% and sideoverlap 75%, must do the job.

Remember I’m not an expert, but I will try to help.
I think you can try these settings and see if results are good enough for you.

orthophoto-resolution = 2.5 to 3.5 (depending on the GSD)
texturing-nadir-weight = 16 to 24
depthmap-resolution = 650 to 750
mesh-octree-depth and = 9 to 10
mesh-size = 100000 to 150000

You may also want to play with texturing-data-term and texturing-tone-mapping

Do some testing and decide whats best for your project.

Regards

2 Likes

@israelbar thanks for the feedback. I will keep it in mind.

1 Like

@javicole4 OK, no gaps is a use-able metric there. You’d also need to define what ‘good resolution’ means :slight_smile:

Have you read these:
https://www.opendronemap.org/2019/09/self-calibration-of-cameras-from-drone-flights/ (there are 3 posts in the series)

…if not, they’re fantastic resources! If you have, and they didn’t answer your questions, it’d be great to hear about what you thought was missing.

For 5ha and trees, you also might want to just collect more imagery than an ‘optimal minimum’ - if you’re not short of batteries, a second lower grid (diagonal crossed) can’t hurt.

Yes, it increases time to process the data. However, the case comes up so often that we think ‘oh we’ll go collect more data later’ and it never happens - so my best advice is to run the camera as hot as possible always whenever you get to the field and worry about the problem of more data than you need rather than the other way around :wink:

Organise flights so that you have a separate set for ’standard grid’, ‘diagonal grid’ and ‘extra detail shots’. So you can start by processing a minimal set, and images as you need to…

4 Likes

@adamsteer Thanks for the feedback. I’m going to read that post and will share my observations with you.
Regards

2 Likes