What can I do with weed mapping?

[matttargett4]

So I’ve been spraying a lot lately, and it struck me that I’m spraying lots of ground that just doesn’t need it so I was thinking of mapping.
i realise that we are about to have a new agricultural revolution and teenagers running crowd funded business which don’t come up with much are going to save the world but in reality what can I do right now?

i am/ was keen on the idea of skippy scout however a bit of a go with it shows that scanning a whole field at a resolution able to identify weeds isn’t practical, and then you need precise location data.

so I was thinking of staying manual, if I was to get one of the Emlid rtk receivers that will work with n trip from my phone, could I log locations and save them in a file format that I could open with say gatekeeper and then export the field map to my auto section system so I drive across the field and the sprayer only turns on where I have logged a weed or area?

is there some better way?

its sounds labour intensive, and it is, but at the same time the herbicides I’ve been spraying can be 10-40 quid per ha, if I could reduce the use it would justify quite a few hours of labour. With the exception of the Emlid receiver I have everything else I’d just need to get it together.

 

[Northern farmer]

Me too been thinking about the same thing. Don't have any solutions but I was wondering what did you mean with manual mapping. Would you walk through the field or did you mean recording weed spots while working at the field for other tasks? In any case actually an Ardusimple simpleRTK2b would be easier to carry and and order of magnitude cheaper.

I don't know Gatekeeper and have only tried Farmworks but with the latter one can draw prescription maps and 0%/100% prescription maps seemed an option to our Amazone sprayer rather than trying to build a coverage map (that excludes the spots to be sprayed).

There was a resent thread here about field mapping where mapping from drone photos was considered easy and accurate. I did not get all the answers but I cannot see why a drone could not take photos at low enough elevation to find weed spots? One can do photo stitching and probably with a couple of RTK accuracy reference points the whole field can be shown on a picture with cm accurate position data. But which one is easier, creating records on the field based on visual observations or post-processing aerial photos. Are skippy scout and similar not able to run object recognition on the drone photos to automate the weed detection process?

A pity that research institutes seem to lead a bit too far to the future to be of help with current issues (I accept that being their task rather than implementing something based on existing knowledge). You can find some studies from sprayer companies web sites like Amazone but I find them a bit of marketing so far. Companies want to show that they can handle the most recent technology but the actual complex product comes years later. Instead they could build something simple that would be useful as of today. I'm counting more on open source enthusiasts starting to build something.

 

[matttargett4]

So when I was talking about manual mapping what I meant was you carry a pole with a receiver across the field with you, when you get a weed you put the pole on itvand store it’s location, ideally with what it was, ideal scenario is a preset of weeds on 5 buttons, 1 for bg, 2 for cleavers etc, you walk around the field, put the staff on the weed and press the relevant button.
you then export this to gatekeeper which shows you the field and then weed locations and types within it and at that point you decide, actually this is a whole field job or it’s a spot spraying job. If it’s spot spraying each weed has a circle drawn around it and that’s a spray zone, map downloaded to sprayer, drive across the field and the sections come on in the relevant places.

as interesting and useful as the drone stuff is it appears flawed to me, I’ve hopefully attached 2 pictures, not actually taken with skippy but that’s not important, the rows are 6 inch so gives an idea of how large an area you can survey with one picture and have good enough resolution to get useful info from it, clearly the thistle is easy to I’d, but the picture is about 1m square. The rosebay willow herb in the other picture is very difficult to I’d, so at 10,000 images per ha the collection is a massive task and the processing even more so.

i have average 4ha fields and what drone flying has taught me is I have to do fields individually and actually I can cover a fair bit of ground on foot in the set up and shut down time, I think where it might work is a preliminary survey.

i was looking for something practical to compliment manual field walking which could perhaps halve my herbicide use and use most of what i have rather than hanging about waiting for what just seem like trendy words to turn into real, affordable solutions.

 

[Northern farmer]

You have considered this much more than myself but my thoughts are very much the same on the parts that I also have considered.

I believe image processing (pattern recognition) is about there (my friend says it is already here but I have not tried it with him) and would solve the huge task to analyse the huge number of photos. Then when image processing works reliably, we still need some SW to produce perhaps shape files that the sprayer understands. Sounds attractive but of no use while waiting for it.

I have hardly any programming skills but even I could take my ArdusimpleRTK2B board with the antenna on a pole (I have it on a wood stick for manual position logs), attach it to my Raspberry pi and build a five button switchboard. Log a position at each button press and attach a number to each position indicating the size of the weed spot. I'm not sure if my programming skills are enough to post-process the file into a shape file that consists of circles described by the log but I know there are many members even at this forum that could do it easily. Gatekeeper should accept that shape file and I believe you know how it goes on from there.

Perhaps I see one detail differently, not sure. If a field only has a few weed spots to be covered, I'm sure simple manual logging is fine. But if there are many spots randomly all over, I'm afraid one can easily miss a few if not walking consistently throughout the field. This is why I've assumed an auto-steered vehicle (preferably smaller than a tractor) to run through the field, say A-B lines at a distance of the largest circle to be logged. In this case logging could also be modified to five "sections" along the A-B line, each section to be sprayed as long as the corresponding button is being pressed.

 

[radu]

i don't think it's feasible walking the fields to record weed locations. For once, it's very easy to miss some of them and each will turn into tens or hundreds by next year. Or there's a rain from the time you record the weeds to when you actually spray and more weeds emerge. Also it would be very time consuming. How many acres a day you reckon you could map? Then you need a big capital investment in the kit (section control & variable rate, handheld precision GPS, data processing) that wouldn't generate the expected results.

I think weed mapping is still a playing ground for the few enthusiasts that have education in both it and agriculture and not lastly, have the time

 

[Northern farmer]

i don't think it's feasible walking the fields to record weed locations. For once, it's very easy to miss some of them and each will turn into tens or hundreds by next year. Or there's a rain from the time you record the weeds to when you actually spray and more weeds emerge. Also it would be very time consuming. How many acres a day you reckon you could map? Then you need a big capital investment in the kit (section control & variable rate, handheld precision GPS, data processing) that wouldn't generate the expected results.

I think weed mapping is still a playing ground for the few enthusiasts that have education in both it and agriculture and not lastly, have the time

I agree on many of your points but I don't get your point on rain. If you spray the whole field, you will not get rid of the weed that comes after rain anyway. Not good but the point of saving chemicals would still apply.

The capital thing I don't agree either. I guess individual nozzle control is still rare but section control is not. The majority of sprayers have the capability to spray only mapped areas if someone was just able to define the spray map.

A handheld GNSS receiver at cm level accuracy costs about 211 Euros (Ardusimple or similar).

A bit of enthusiasm is the main missing piece.

 

[Dave79]

If you’re going to walk around the field to mark them with an antenna, would you not be better to just put on a knapsack and spot spray them? Somebody told me about a video online that shows a sprayer spot spraying off a camera mounted on the tractor, as in a sprayer at full width driving up and down field and the nozzles only firing when there is a weed. Neat trick but probably £££

 

[radu]

I agree on many of your points but I don't get your point on rain. If you spray the whole field, you will not get rid of the weed that comes after rain anyway. Not good but the point of saving chemicals would still apply.

The capital thing I don't agree either. I guess individual nozzle control is still rare but section control is not. The majority of sprayers have the capability to spray only mapped areas if someone was just able to define the spray map.

A handheld GNSS receiver at cm level accuracy costs about 211 Euros (Ardusimple or similar).

A bit of enthusiasm is the main missing piece.

You need to apply the herbicide at a certain stage of the crop and when the weather is right. Let's suppose you apply it on D day. If you spray the whole field, you hit all the weeds that are above ground that day. If you were using a map created from data that might be from let's say D - 3, even if the surveyor was very rigurous and didn't miss any weed, in those 3 days between mapping and spraying some weeds might emerge and they won't get hit. So, you save some herbicide but risk not having the best possible efficiency (for that particular day) in weed control. I guess you could resurvey it a week later and make the necessary corrections but the amount of manual labour required is huge. Sure, it would work if that is what one tries to prove but is it feasible?

Besides section control which is pretty straight forward on a sprayer, i think you need vari rate activation and some skills to turn the data from an ardusimple receiver to a compatible shp file. That might not be so common.

 

[Northern farmer]

How do you define efficiency? It is obvious that spraying the whole field results to the least weed left. Efficiency should mean the least weed left for a given amount of chemicals (some countries want limit the use of chemicals as much as possible, not only to save money).

The skills to turn the ardusimple position samples into shapefiles we already discussed. This asks for a very tiny bit of SW skills, not expensive at all. Not everyone has the necessary skills but we've seen other extremely complex applications shared for free, why not this one if we ask politely.

If we only want to find flaws from previous proposals, we see less progress than in the case we try to provide solutions for the issues we find!

 

[radu]

Efficiency = profit margin.

Half of the ideea is very good, spraying only where it's needed. The other half is not so good, walking every field and surveying every weed could turn up to cost more than the savings from using less herbicide.

An alternative way of creating the weed maps, using field images and software might require less effort in the long run and provide better results but i don't think there is an off the shelf solution that does this.

However, it's easy to judge from thousands of miles away, from an armchair. The best thing would be for the OP would try it his way and give us feedback. I too recommend Ardusimple, it's cheap and precise.

 

[matttargett4]

I think we may be looking from different angles, I’m not sure how the world is in Romania but here profit margin certainly isn’t the only measure to consider, there is nobody who thinks pesticides are good, we are dumping our produce currently in favour of cheap imports, our choices are clear, use less or none.

on the economic front, i can spend 3 man hrs a hectare for the value of some sprays, 3hrs would do a heap of checking.

i absolutely get the argument for automation, it’s correct, but I have been looking a bit more, I’ve signed up for a trial of drone deploy, it appears if I spend £100/month on it, 6k on a special camera and another subscription on software there’s a chance it can be done with a dose of manual input, it is as you suggest for those who have time education and money.

i thought maybe drone deploy could take pictures and I could look at them all and decide on a grid system to spray or not to, trouble is, drone deploy min altitude gives a picture covering 10m square giving a supposed resolution of 0.25” per pixel, I defy any Person or computer to find a weed with that. Even then 5 acs takes 20mins to fly and there are 270 images to analyse.

so as far as I can see at the moment manual is my only option.

i think I need a grid system, say 4m square, walk Perpendicular to tramlines back and forth at 4m spacing, deciding yes or no on every square.

perhaps I will try a few acres to see how long it takes

 

[radu]

in my mind, the goal of a business is to make money (and needs to have a viable business plan on the longish term). however, our personal goals should definitely not be that narrow, so i agree with you that on a personal level profit shouldn't be the top priority.

The grid system is a good idea, no need to manually pinpoint every weed in the field. Have you thought about how you will implement this grid? I have some ideas... but for the beginning, i think it would be best to keep it as uncomplicated as possible.

Good luck with this endeavor and keep us updated!

 

[aidan]

I think we may be looking from different angles, I’m not sure how the world is in Romania but here profit margin certainly isn’t the only measure to consider, there is nobody who thinks pesticides are good, we are dumping our produce currently in favour of cheap imports, our choices are clear, use less or none.

on the economic front, i can spend 3 man hrs a hectare for the value of some sprays, 3hrs would do a heap of checking.

i absolutely get the argument for automation, it’s correct, but I have been looking a bit more, I’ve signed up for a trial of drone deploy, it appears if I spend £100/month on it, 6k on a special camera and another subscription on software there’s a chance it can be done with a dose of manual input, it is as you suggest for those who have time education and money.

i thought maybe drone deploy could take pictures and I could look at them all and decide on a grid system to spray or not to, trouble is, drone deploy min altitude gives a picture covering 10m square giving a supposed resolution of 0.25” per pixel, I defy any Person or computer to find a weed with that. Even then 5 acs takes 20mins to fly and there are 270 images to analyse.

so as far as I can see at the moment manual is my only option.

i think I need a grid system, say 4m square, walk Perpendicular to tramlines back and forth at 4m spacing, deciding yes or no on every square.

perhaps I will try a few acres to see how long it takes

What type of camera is it for 6k

 

[SoilEssentials]

Here at SoilEssentials (www.soilessentials.com) have experience in mapping weeds, providing drone services and providing a cloud platform to handle your data along with drone imagery. We have some relevant projects at the moment towards weed mapping and in particular reduction in chemical use and therefore environmental and cost savings. We are a year into a 3 year roadmap to develop a commercial product to detect and spray only weeds in grassland. The project is called SKAi, the SoilEssentials KORE Artificial Intelligence platform, this is a retrainable, smart-camera, vision system for agriculture. This means that one of the unique properties is we can train our machine models with imagery (your own if need be) from a drone, mobile phone camera or using our scouting app in KORE (www.koresolution.com)

We are undergoing prototype trials just now and over the summer, so feel free to get in touch and start the conversation.

 

[farmerm]

small drone company is doing ground level weed mapping with the aim of then using small 4x4 drones to treat individual weeds.

One thing to feed into the discussion... Ag Chem companies need sales volumes to produce product and invest in R&D. If we are all significantly reduced the amounts of chemicals we apply it will either have to become vastly more expensive or they will walk away from the sector and we wont see any new chemistry come though.

 

[matttargett4]

Having spoken to Jim Wilson from soil essentials last week about what they are doing it’s very interesting, however it’s clear from what they’ve found it’s not as simple as a silver bullet, we are going to try a few fields on a grid system scouted manually, given the time of yr they will likely be maize, will be interesting to see what comes out of it.

i took a look at the small robot company website again this evening, great but it still isn’t actually a working product or system which I can buy yet. I also wonder how it fits in an mixed farm spread over 8 or 10 miles?
something else which has come to my attention by chance this week is camera guided hoeing of cereals, having loosely thought about the economics of it I think it’s something I’m going to investigate further.
a crop hoed a couple of times in season with glyphosate out of season to control perennials could be a possibility and in a resistant bg location spending £100/ha on herbicides, or a rg situation might stack up financially and have no resistance issues.

 

[DanniAgro]

small drone company is doing ground level weed mapping with the aim of then using small 4x4 drones to treat individual weeds.

One thing to feed into the discussion... Ag Chem companies need sales volumes to produce product and invest in R&D. If we are all significantly reduced the amounts of chemicals we apply it will either have to become vastly more expensive or they will walk away from the sector and we wont see any new chemistry come though.

Yes, they will probably abandon the market.

 

[PSQ]

As soon as you start hearing ‘artificial intelligence’ and drones etc, you know that it’s going to cost considerably more financially and in management time, than spraying at a flat rate across the field.

For £50 you can buy dozens of 4mm glass fibre canes and a smartphone GPS app (I use Waypoint, think it was free) to map weedy areas. The following spring the canes go out, leaning them towards the patches to be sprayed. You just need to be creative with your file names in that it’s easy to get muddled up, so I categorise by field, year, tramline or other identifiable mark, and point number: ‘Bruntrigs 2017 T4 1/4’. In Waypoint it has a compass icon to direct you to previously saved points.
Simple, repeatable, effective and doesn’t require £5,000 to buy into some ineffective and ‘soon to be forgotten’ half baked technological wet dream.

 

[SoilEssentials]

Being farmers our self any product we bring to market has to be affordable and overall, make good business sense.
New technology doesn’t have to be expensive....but it has to work!

 

[DanniAgro]

Being farmers our self any product we bring to market has to be affordable and overall, make good business sense.
New technology doesn’t have to be expensive....but it has to work!

It doesn't have to be expensive, no, but unfortunately it has a nasty habit of becoming so.