Is there somebody who doing agro inspections and using ODM for stitching images?
I would like to start a business with agroinspection and I have a following issue:
I make my first field overflight approx 80 ha. For testing i use fieldagent from sentera and i set recommend settings (altitude, overlap, speed) like you can see on below picture.
Problem was when i wanted to stitch images in WEB ODM. It stitched only couple of images not all images. After study i found out that overlap have to be min 80%.
But problem is, when i set 80% overlap it will be so dense drone track that i will need 9 battery and up to 3700 images. And this is only 60 ha. I have for example potencional customer who have up to 1000 ha.
So question is:
- Is possible to make images stitching in ODM with less overlap like 80% (what settings you recommend for it?)
- Or how professionals make overflight big fields?
Thank you for your advices.
PS.: You know i am beginner, so sorry for basic questions
For an agricultural field, you may be able to get away with as low as 60% overlap and sidelap. If it’s pretty flat, you can try adding the
for me it strange that you have 2.6cm/px from 92m altitude,
last time i flew my phantom 3 on 65-70m and i had 2.7cm/px resolution.
btw. try 70% overlap and if you don’t need a high gnss accuracy then try increase speed to 7-8m/s (if you have sunny day - remember that shutter speed should not be slowest then 1/250s)
or split you fly for few smaller mission, it always work for me, and it’s more secure - you can backup pic after each landing.
My experience on my own place is that the 80% is needed to get enough matches, though you may get away with slightly less. I normally run at 5 m/s and 92m with a P4P. I set up around 30-32 ha per run which uses my 3 batteries up (about 60 min on-task). With my RGN camera attached this is reduced to around 20 ha (only about 45 min on-task) - but that is OK for me as the fields I am interested in getting NDVI are much smaller. Your number of photos seems really high - I tend to get around 700-800 per run (with the DJI 20Mpx camera). This gives me a similar total number of photos (3800-4000) for all 160ha (over 5 runs).
Part of the issue here is simple maths using a drone. Even if you are happy with more than 2.5cm/px, in most countries your are limited to 120m altitude, which will still constrain the area captured in each photo. The relatively short life of the batteries is also an issue. You can probably increase the drone speed a little, but that will also depend on the ambient light conditions (as you will need higher shutter speeds) but it may also be a further battery drain. So 120m, 8-10m/s, 70-75% overlap may be achievable depending on your exact requirements.
I’d be thinking that if you are looking at 1000ha it may be possible over a week using 12 or so batteries with a P4P (say 4 per 80 min sortie , 3 per day moving base between sorties). The other possibility being the use of a fixed wing drone. Either way you are probably in for a significant extra investment (and only you can tell whether it is cost effective).
Not really true for 60% side/overlap !
Depending on sensor… Some sensors like parrot sequoia or micasense rededge have only a small size of good values and it’s on center of frame : cheap sensor, poor lenses results in high or reverse vignetting. High side/overlap compensate this problem.
i use a mapir camera (survey 3W ocn) for biomass estimation on my farm. I also started with a copter, but even with 35 min flight time the maximum field size is limited to 30 ha. In my country copters are limited to 100m and with the mapir camera I have to maintain 80% over/sidelap because of vignetting.
At the moment i use a fixed wing flying at 200m with 16 m/s covering over 300 ha in one flight (45 min), what is very important because of the changing light conditions which significantly affect the biomass estimation.
In WEBODM i use the fast-orthophoto option with no resize and min-num-features at 12000. Processing takes about 2h for a 150ha field with 450 pics (1000,-€ gaming PC).
The results are good enough for estimating zones of different biomass for creating application maps.
Welcome Hubertus! This is a really cool overview. Thank you for sharing.