Hello from Brisbane, Australia

[farmguy1]

Hi All;

I am a farm owner with a programming background.

I’ve just joined the community and have WebODM setup on my linux box. I’m very happy to say that the setup went without a hitch. Also loving that my testing of images from my previous DJI Phantom flights from back in 2018 using Drone Deploy look great in the WebODM interface.

I’m here because I’m planning to map two farming properties as part of a carbon sequestration project using a Cessna and a GoPro clone.

I have read about the challenges with GoPro’s here and once I get my thoughts assembled I was hoping to get some general comment on my plans.

Thanks for the information on this site and I look forward to using WebODM in my project.

Cheers,

Phil

[pierotofy]

Hi, welcome!

The software fully supports fish-eye lenses (GoPros included), just make sure to process the images with the correct lens model (via the camera-lens task option).

[farmguy1]

Thanks!

[Gordon]

Last year I mapped a ~1400ha property which had been planted out with something like 50000 trees for Carbon sequestration, with my M2P. With a GSD of 2.3cm, it took over 23000 images!

[farmguy1]

Yowza…

[Gordon]

You might find some useful info in this topic, where I go into some of what was required for the job above:

Avoiding Bowling over large areas

Smathermather’s Weblog has some great insights into avoiding bowling effects when imaging over large areas, and with a recently completed job I incorporated some of the methods to minimise bowling. The mission was to create a baseline ‘before’ image of a large (~1400Ha) property that is being re-forested, to bring it back to a state similar to how it was before clearing and decades of livestock grazing, as a Carbon farming project. The land is situated in an ancient volcanic province, somethin…

[farmguy1]

Thanks Gordon. That’s very helpful. It sounds like a great project. So much data to process.

I guess the 23000 images was due to the accuracy that you achieved. I’m only looking for half a meter accuracy and hopefully a process that is easy for my pilot to manage so I don’t have to be there next time.

I will see if I can incorporate all Smathermather’s points into my flight plan.

Thanks again.

[Saijin_Naib]

Welcome!

Sounds like an amazing project!

What GoPros are you planning on using?

The main issue is that we don’t currently correct for Rolling Shutter Distortion, but from what I gather, the most recent GoPros have a “fast enough” read-out that other suites don’t treat them as such (though at Cessna speeds, they may need to be still treated that way).

[farmguy1]

Hi and thanks everyone for your interest. I’ve gathered my thoughts and am putting them down here. It’s a fair bit of detail straight up, I figured I may as well just get it all out there.

As mentioned I am planning a flyover of two properties to develop a carbon farming project. The site is 2000 Ha so I was looking at a Cessna flyover with a GoPro clone. The terrain is relatively flat with a couple of hills and a number of creeks and gullies.

I know after my reading of the posts and comments here that there is a rolling shutter problem with cheap action cameras, limited overlap and a moving camera. However I’m wondering if the problem can be resolved using the 2 layer flight plan from following recommendations from this post (Optimizing flight planning for calibration (or Self Calibration of Cameras from Drone Flights, part 4) – Smathermather's Weblog).

The approach would be to attach a 20 MP Campark X30 GoPro clone under the wing with a triple suction cup mount.

The camera can be controlled from the cockpit via an Android App to take 10 sec time delayed images. The resolution would be reduced from the 20 MP max to 13 MP.

I was planning to update the GPS data back in the office from a Android phone running the GPSLogger app (3m accuracy). The phone would be time synced with the camera and located on the dash of the plane . A consumer grade Garmin GPS would also be on the dash in order to have a cross check/backup. I have the option of setting up GCP’s.

The flight would be on a bright sunny day timed for midday at 120 km/hour (33m/sec). At this stage, if I get a clear ortho image with +/- .5m accuracy I’ll be very happy on my first run.

I’ve configured QGIS Flight Planner with the cameras current settings with a 30% frontlap, 60% sidelap, and a camera tilt of 80 degrees. The direction of the second layer of images would be offset by 20%.

This plan can be easily changed if needed. e.g. 80% sidelap. The Layer 2 config is shown below.

The two output layers from QGIS Flight Planner look like;

     Layer1(Pink); requires a minimum of 95 images 1410m apart at ASL 903m.
     Layer2(Red); requires a minimum of  571 images 475m apart at ASL 767m.

The best settings of the camera that I’ve found from taking photos in full sun are; Photo Resolution = 13M, distortion correction = On, ISO = 100, White Balance = Sunny, Exposure = 0, Sharpness = High , FOV = Narrow

Note: These settings translate to the following on the images properties;

     Width: 4160 Pixels
     Height: 3120 Pixels
     Aperture f/2.6
     Exposure time 1/3120sec
     Focal Length 18.0(35mm film), 2.3(lens)
     Sensor Size 8.46
     Image size is 5 MB

This project will be run on a shoestring budget and the question at the moment can be boiled down to go/no-go. If its “go” then there are two additional questions in my mind;

1. Are there better/recommended settings for this type of camera to work with WebODM?
2. Can this flight plan be improved to overcome the rolling shutter issues at 120km per hr?

Thanks again for your consideration.

[Saijin_Naib]

Quickly, I would not recommend going below 65% for overlap or sidelap.

[Gordon]

@farmguy1

Aperture f/2.6
     Exposure time 1/3120sec

The sharpest images, especially around the edges are generally obtained when the camera is stopped down a couple of stops, the high plane speed isn’t an issue for motion blur with such a large GSD, so f/4 or f/5.6 should produce sharper images, so long as the shutter speed is faster than 1/125 sec.

See: Incomplete Orthomosaic - #17 by Gordon

BTW, I recognise where this is, having bushwalked and cycled around the area quite a few times, before they dug a monstrous hole for coal.

Will the pilot be terrain following? The overlap will be reduced over higher ground if flown at constant hight.

[farmguy1]

Thanks Gordon. I am amazed that you could recognise the area from such a small bit of satellite imagery. I guess the left hand corner gave it away.

That’s very helpful re plane speed. We will fly at a set altitude and I’ll increase my overlap as per the comment from Saijin.

I was of the view that the aperture is fixed on a GoPro type camera. I’ll look into tweaking the ISO and exposure settings and see where that leads.

[Gordon]

@farmguy1

a camera tilt of 80 degrees.

Set for a flight speed of 120km/hr rather than parked on the ground, but if there is a head/tail wind blowing of any significance, will the pitch of the plane will vary, varying your set 80°? Cross wind will cause a yaw variation, so probably best to increase your overlaps over what is essential.

@farmguy1

I was of the view that the aperture is fixed on a GoPro type camera.

I know it is on my cheap GoPro copy, but wasn’t sure about the camera you are using.

One other thought - does setting sharpness to ‘high’ introduce any artefacts on the images at all with that camera? I’m thinking rings around small objects etc, which I often see in sharpened photos.

@farmguy1

I am amazed that you could recognise the area from such a small bit of satellite imagery.

Actually, I just saw the email notification, which had your name My wife and I visited the FLAC camp which I think was on your property, that must have been 8 or so years ago.

So let me know how it goes in case I can help out in any way!

[Johnny5]

Welcome Phil,

I’m down in Sydney and currently learning WebODM, with a goal of figuring out how best to process larger than average batches of photos (say 3000+). If you’re ok with sharing photos with me I can help out with running a few tests to help figure out optimal processing configuration (I’ve got a server with 512GB RAM, 2x20Core/40 thread Xeon CPUs and a 12GB GPU).

Cheers, Johnny

[farmguy1]

Thanks Johnny. Your machine sounds way over anything I have. I’ll be able to share for the purposes of your testing but I would appreciate it if you did not share the images or the results of the WebODM output elsewhere.

The flyover is scheduled for mid to late May. Hopefully by then I will have refined my flight plan and camera settings.

[farmguy1]

Thanks again for your comments Gordon and your support of FLAC. FLAC was on Cliff Wallace’s place next door. Time certainly has got away.

Sounds like 80% overlap (sidelap?) is the go. I’ll check for rings on the sharpened setting. I also wonder about the merits of the “Distortion Correction” and whether ODM does this anyway in post processing.

[Johnny5]

No worries Phil, yes I’ll keep it confidential.