Hi there, I’m a broad acre farmer in Australia growing wheat and lentils. It looks like DJI enterprise drones are doing a good job at picking up weeds but I’m wondering how small of a weed they successfully detect? (Coin, golf ball size… ect.)
Ideally I would love to use this tech for targeting hard to kill weeds in summer, if it could I would also map in crop with ndvi and look at any hot zones that are high in biomass early on before canopy closure… with the assumption that these areas are weeds.
I already have a sprayer RTK ready for this sort of thing with John Deere operations centre.
Any feedback would be appreciated, bonus points for farmer feedback in my situation.
Cheers.
the size of the weeds to be detected depends on the sensor and the flight height above ground. The higher the resolution of your sensor the higher you can fly to detect weeds at the same size or smaller size when you fly lower.
I will give you some examples for the most common drone which is the Mavic 3 Multispectral:
Flown at ~ 110m, each pixel represents 5cm on the ground. So it has a GSD of 5cm. Thats Sorghum in Corn V4.
This is the multispectral sensor of the M3M which has a resolution of 5 Mega Pixels.
Here you have another comparison with the RGB sensor of the M3M which is 20 Mega Pixel at what to expect at different heights:
What you aim to achieve is very well possible. You can also use the NDVI to do pixel based analysis, if you you understand well what weeds are and what not. This can all be rolled over into accurate prescription maps:
It looks like RGB @ a 12m flight height is going to give the best chance at picking up the smallest of weeds. Summer spraying is green targets on brown so that should make things easier Im guessing… Am I correct in thinking that?
How many hectares and hour could be covered at this height? Being a broadacre farmer my paddocks range from 40-200ha in size.
Assuming you get a Mavic 3 Multispectral here is a table comparing mapping speeds and ground sample resolutions at different flight heights. It also makes a difference if you fly only RGB or RGB+MSP as the Multispectral sensor is slower:
Based on this data you will see, that mapping at 12m height is unfeasible for practical field sizes like you mentioned. Atleast thats the case for the mentioned DJI M3M , if you for example have a 60MP sensor on a another drone it looks completly different, as you can fly higher and cover more hectares while maintaining good ground resolution.
Now some practical adivice:
For Green-on-Brown targtes (assuming every green plant on your field is a target and the rest is soil or fallow) I would fly between 70m and 110m and use the multispectral imagery. Because with that you can create very sensitive vegetation indices lile the NDVI. Even though you will not be able to ID the weeds or volunteers , you will know where they are! With PIX4Dfields you can then create a map which selects only parts where “green” pixels are.
@ 70m you would be able to map 100 hectares in 2h with multispectral
@ 90m you would be able to map 100 hectares in 1,3h with multispectral
@ 110m you would be able to map 100 hectares in 0,8h with multispectral
If you only fly in RGB which also works, but then you are limited to VARI and TGI indices which are a bit less sensitive (so you might miss some small areas) you could already map @ 90m with full speed of 100 hectares in 0.5h.
Here you can see how such a workflow looks like in PIX4Dfields, and that was based on RGB imagery only:
Thanks for the information, all the right stuff to help me figure out if this will be cost effective and practical for my scale.
Do you know if farmers are doing this on my scale? May also be worth noting my total farm size is 1200ha. Meaning if it takes 4 hours (or more) to do 100ha so be it…, not a huge problem would just mean going through a few battery’s. Quality of job comes first.
You will find the best flight settings according to your crops and type of weeds after a few flights I would say. We know companies that are flying and processing 1000ha a day flying a M3M higher than 120mts, in fact is the example case that Julius mentioned of Sorghum in Corn V4:
Yeah there a guys who doing it in australia already with PIX4Dfields at scale, unfortunatly I have no direct contacts I could put you in contact with.
Like Jose said, you will quickly find the gist of it! Some more practical advise:
Invest in a big sd card for your drone which are also fast 256GB - 512GB (like SanDisk Extreme PRO SDXC UHS-I 256 GB Memory Card), buy 2 of those and you are covered. You will find, that copying data from the card to you computer with PIX4Dfields is the bottleneck you want to get rid off.
With 4 batterys and a Powerstation you can fly non stop, maybe even with 3 batteries.
Try if RGB only flights are good enough for what you want to achive. As the drone can fly much much faster in RGB only.
So I went ahead and got myself setup with the DJI Mavic3 multispectural and looking forward to giving some user feedback once I’m up and running in the sprayer.
Ive run into some issues with sending a prescription to John Deere operations Center but somehow managed to get a file in there, however it created its own new field… I archived the new field and kept the prescription file and made a work order with it in the correct field and all is good there now. (Still having trouble sending operation files)
Now in the sprayer I have my shp file loaded up correctly. I’m finding my sprayer does a really good job at fireing different sections when there is still at least 1 going.
It has a hard time keeping up when only 1 section needs to fire here and there and the whole boom is off between areas to be sprayed.
I will try to upload a video for reference, see how I go… it wouldn’t simply let me do this from my iPhone.
This could obviously be more of an issue with my sprayer or its software also? It’s like it’s lagging and past the point it needed to spray.
However it is timed perfectly when multiply sections going on and off consistently.
My sprayer is a SP John Deere R4060 with 11 sections (3.3m each)
I mapped these using ndvi @ 90m which created good maps using pix4d fields.
These maps were done without RTK (now have that working)
my grid size is set to 3.3m to match my sprayer sections.
pix4d is authorized in JD ops center, using the share export I am able to get layers into there. The operation files not as easily as mentioned above. Exporting a shp file manually to my desktop then into an Rx folder to JD ops center was successful but the map was not what it should be… it had a red line through the paddock with some odd symbols. This could be the problem…?
I am already using JD ops center to document all data so i already have all my fields, boundary’s, A-B lines ect setup.
I have watched the video all about the john deere intergration.
EDIT*** I have managed to get the prescription maps from pix4d to ops center now. However I have had to get them there by selecting the organization, client, farm and paddock that it automatically created in the ops center. So the file does end up in my files page.
Is this how it should work? I do have the option to select my actual organization, client, farm, field… but the files wont show up in JD Operation center doing this.
sorry for the long text here. Just trying to get as much info to you as I can. Our time zones obviously don’t match and I normally get a reply around 2am here haha
hi brayden,
I’m in a similar situation as you where and debating which way I should go, either a fixed wing or the M3M, are you still happy with how many Ha a day you can do with the M3M, and how many ha and at what height can you get from one battery
Hey there. I’m happy with the area/hour I’m getting but it comes back to the size of your farm and what you consider to be reasonable.
I’ve been doing in crop RGB camera only looking for patches of grass in wheat. At 90-110m this is 200ha or so an hour. I have made an adjustment to the overlap setting and down to 55% not 70% (maybe someone can comment if there may be issues with adjusting that? Still makes a map in pix4d fields)
However summer weeds I found NDVI much more accurate and easy. 90m is a good height to start and I think this is around 70 ha/hour
110m was also just as good where I did it. (90ha/hr maybe)
Also have gone down to 70m where I wanted peace of mind that I got every weed. (About 50 ha/hour)
If I change the overlap to 55% these numbers would jump up which I’ll try.
Battery’s last about 35-40 mins. It depends on flight speed and wind. Also keep in mind how far the paddock is from where the drone takes off from as more travel distance = less time mapping
There’s a good pic in this thread I think. It gives you an idea of speeds at different heights
Hello @bjh_rox ! Your experience is very interesting. Im also a broadacre acre from Argentina working around 1500 hectares with summer crops. Corn and soybeans rotation. Just got my Mavic 3m and starting to experiment with Pix4d. I have two main questions for you! Im looking to do green on brown spot spraying with ground sprayers.
First question: Are you getting good enough results only with big patches of weeds or also spraying small isolated weeds around de field?
Second question: How big of a diference are you noticing using RTK? I dont have an RTK station at the moment.
Hey there, I was getting isolated weeds in summer fallow. Make sure you’re using NDVI for best accuracy.
RTK is a must and I’m convinced after a flight I did only last week, I did a flight without it and the drone missed its home landing pad by a few feet.
Thanks for the great feed! I was worried that sprying isolated weeds would not be an option and only patches of weeds could sprayed. And thanks for the advice about ndvi for accuracy, totally agree on that
decreasing the overlap will increase the number of images required
120 m
80/80 overlap
rgb +msp
will work just fine
also link John deere in the share menue at the top left of export menue
enable rtk to your local gnns network it should be a free council service in aus
We - and our partners - use cookies to deliver our services. By using our website, you agree to the use of cookies as described in our
Cookie Policy
These cookies are necessary for the website to function and cannot be switched off in our systems.
They are usually only set in response to actions made by you which amount to a request for services, such as setting your privacy preferences,
logging in, or filling in forms. These cookies do not store any personally identifiable information.
These cookies allow us to count visits and traffic sources so we can measure and improve the performance of our site.
They help us to know which pages are the most and least popular and see how visitors move around the site.
All information these cookies collect is aggregated and therefore anonymous.
If you do not allow these cookies we will not know when you have visited our site, and will not be able to monitor its performance.
These cookies may be set through our site by our advertising partner (Google).
They may be used by Google to build a profile of your interests and show you relevant adverts on other sites.
They do not directly store personal information but are based on uniquely identifying your browser and internet device.
If you do not allow these cookies, you will experience less targeted advertising.