How to efficiently perform agriculture based drone missions to identify crop species and densities
We are hey everyone and welcome to another episode of ask drone. You got a head spinner for you today. As always, my name is paul. My name is rob super excited to be here with you, as always, and very much appreciative that youre here with us. Yes, definitely appreciative, as always, i just want to say a quick shout out to pix4d that cyber monday deal they had on software man. That was a that was a deal, and i think a lot of people appreciate that. So, anyway, lets get right into todays question, which is all about ag agriculture and someone in portugal trying to solve a problem. I do. I do rob, feel a little bit of connection and relevancy to this caller. How so because i have a portuguese water dog. Oh Laughter, oh, that got teed up perfectly. Yes, it did kona. We would love to go, see your motherland. I want to go there really badly. Well, anyway, lets get into that question hello. My name is uh vashq. I am planning to start a business over in portugal and my question, for you is, if you guys know some software, that you could recommend to process some agricultural photos. We have rice plantations that have various infestations with four different types of plants, and i was wondering if the software or, if you know any software that can differentiate between the rice and some of these plants, which are very similar to the rice itself.
The four different species are equinochloa cruz, galli, ciprus, maritimus cipros deformis et. Thank you very much, and i appreciate your help. Thank you. That might be the most detailed question weve ever had in terms of actually naming the the plants wow um, yeah paul. This is all you askdroneyou.com by the way is where he asked his question, because we we know that a lot of you out there have questions as well. So please get them in wed love to hear from you, but man. What do you? I got nothing. Well, i mean without having uh the empirical experience, which i just want to preface the show by saying that were gon na kind of run through some systems that we know um. That may help the caller, but without having tried it ourselves. I kind of always like warn like hey: i havent actually done this um, but if i were to try it heres kind of how i would go about doing it, um and so first things. First um. We know that plants and materials as a whole absorb and reflect various bands of light. Okay, and until we have hyperspectral cameras on drones, it may be quite difficult to be able to isolate any specific plant from another one aka the whole uh needle in a haystack idea, right, yeah lets, say: uh robs wife jana lets go lower than that um lets Say jakes new wife drops her beautiful turquoise ring near the surface of a haystack okay to our eyes, probably hard to find.
But in the eyes of a hyperspectral sensor, we can set up the sensor to look just for turquoise based off of the light it reflects and the light it absorbs. So essentially, everything in the image will be blacked out, except for that one piece of turquoise. Okay, unfortunately, we dont have that capacity with most drones today, unless you are in the military field, so that said or have 50 grand, i would say, a million. Oh those sensors have to be water, cooled so youre. Looking at a big bird to carry that heavy payload, so okay, but that said, lets lets, imagine that this may be possible with a multi spectral sensor. So im going to go through two scenarios scenario: number one is that you just are using whatever mapping drone that you normally use for ag lets just say its a digital camera, though the eb, the wingtra again just a regular digital camera, no special bands of light. First thing is: first: if there is a color deviation between these plants, maybe you could tell the difference in doing a map now. The things to note, though, is that as a passive sensor, meaning were not shooting anything at these uh plants. Our capacity to interpret the map will only be as good as the images that we take, meaning we might have to fly lower to actually identify any deviations of plant species. Okay, so thats number one. I kind of wanted to lay that foundation before we went into multispectral, so lets say that hes got an eb or a wingtra and the um red edge, xm camera right, multi, spectral, camera all global shutters.
We cannotate various bands of light rgb uh near infrared um. I forget what the r d mean im missing two um anyway long story short is flying again on the assumption that were flying low enough to be able to see the deviations in plant species. Would the different plants show up differently in a multi spectral sensor, meaning and the way to test it was? He could just essentially fly his red edge camera, his multi spectral camera at say 10 or 20 feet where he knows that there are rice and these different species of invasive plants and if you can take that photo and i and just process just a couple of Photos and say pix4d fields if they do show up as different objects or as different colors, then the answer would be. Yes, you could differentiate between those different species of plants, but in all honesty, i honestly do not know if the software is going to be able to deviate between the different plant species. Okay, in my eyes, its going to be dependent on two things. Number one is going to be the type of camera that we use. Number two is going to be the altitude that we fly and number three is going to be the processing options for the software and how we actually run the maps because oftentimes a lot of ag stuff is just a simple ndvi map created from multi spectral, imagery That is not going to showcase a deviation in plant species.
Okay. So what do you got? No im, just looking and yeah. Clearly theres thats, all its not going to be any different exactly you can see. So i think, without like a a hyperspectral camera, it may be very difficult to differentiate between the different plant species but uh with these guys doing exactly what theyre trying to do. I would say you know what i think you need to run some tests just take images with your multi spectral sensor. Can you see differences in the bands of light between those different plant species, because if you can then theres, probably a high propensity that you could actually use something like pix4d fields to produce the maps that youre looking for to identify those invasive species? But until we get more advanced camera payloads that are small enough to fly on drones, this might be one of those things that uh might be difficult and so – and i could also be wrong – i mean again, i dont have practical experience with this. So what about deploying something like locate now, i know in order for that, we come back to theres, going to have to be some color variations, but would something like that work so lets assume that there are color deviations between the different plant species? To answer your question, i would not personally use locate, i would use pictera, i think, its a its a far superior product and because i can train pictera with deep learning to essentially eliminate false positives, its a much more viable product for my business.
Okay – and i wonder, and so thats fine, but its the same concept. Yes, its theyre, both built off of matlab so yeah, how big of a difference or how big of a variation in color would there need to be for pictera to be effective in telling you theres a different plan, thats, a very good question: um dont know i Dont know i wonder, but hold on you bring up something important. I talked about in pre show, but we didnt talk about yet on this show, which is all those tests. I said that to run you know with the multi spectral camera, seeing if its different bands of light, seeing if theres color deviation, i would do it on a bright sunny day, and i would also do it on an overcast day because it makes there is a Huge difference in map quality on an overcast day, and when you have that flat light, it will be a lot easier to discern the differences in color, interesting and so like im like huh. So as you test that test it in bright light and also test it in overcast weather, your results may vary very interesting yeah i mean it would be fun to test and to give all this a shot. Add it to the repertoire of things that need to be done. Lets just lets bring on some more members, scale drone you and i would love to spend two days a week of just trying these kind of edge, outlier uh use cases of drones, because the more and more i learn the more and more.
I learn whats possible. Well and some of them, i dont know that we could even categorize them as edge anymore, because i know kind of seeing all the conversations within our community and so forth and just receiving the feedback and discussions from members and non members alike. Theres, more and more people talking about ag and asking about ag so anyways uh, very interesting, yeah, yeah, yeah, very, very interesting, indeed um. I also know because we have some students down in arkansas who have rice paddies, and i remember the biggest issue with them was actually getting accuracy on uh, the the topos and the elevations, because theyre, essentially using up so much water that the water table is depleting And the earth is captaining literally on their field, so meaning their fields are sinking literally, and i mean you remember that guy well, i think his name was rob uh as well and uh. I remember that he was working with wingtra to be able to get that high level ppk accuracy to be able to. You know thinking about paul, no, not his name with rob really. I can see his face right now, one of my favorite students but uh. I uh interesting yeah, no huh, okay anyways, that thats very interesting. I had no idea, but you know that all said there are a lot of applications when it comes to agriculture and sometimes its going to be test and see if it works, but also youre going to have to act like a scientist, eliminate variables and try different Things to see what works and what doesnt, yeah and thats pretty much what we did to be able to even teach pix4d about some of the things that we learned in complex 3d, modeling to essentially hack their software to make it work better.
So scientific method is alive and well, it sure is, it sure is – and i agree with you too, about a lot of these things are not really edge. Applications anymore were seeing more and more and more of it. You know, i wonder if precision hawk could be it offers a camera payload that would be able to distinguish those different plant species. I mean there are drones out there that might work another one that i would recommend looking at and no one ever talks about them. Yamaha yamaha had an agriculture drone. The year i was born huh interesting, the rmax. They called it the max at the time, or i forget what like r800 or something like that, but i just like the yz250 thats all. I think well uh kid drone you go in on a yz if droney is willing to buy some more miners. I dont see why not so anyway, um but uh, but i mean this is this is kind of the fun of being a drone pilot is truly discerning what works and what doesnt, and you know the fact that hes in a different country and not the united States he has a major major leg up, because the faa enforcement, or not, is really not allowing uh for the exploration and innovation of agricultural based things when you have to get a part 137 certificate to be able to get full utilization from your ag. Drone. To be able to spray yeah the irony, and that is that, in terms of a safe environment, it probably doesnt get any safer, but anywho yeah, im sure theres reasons, and i know the enforcement level is near zero.
But at the same time, i think its important at some point – and this is an article ive – wanted to write for a very long time is at what point does enforcement inhibit or empower innovation and as drone pilots? Should we really be asking how important regulatory enforcement? How important it is because a lack of enforcement empowers innovation, and you know every experienced drone pilot knows exactly what im talking about im, not going to spill the beans on the show but um that lack of enforcement can potentially provide for innovation that the u.s needs To have in order to maintain our economic powerhouse uh of a society that we have does that, if that makes sense yeah, i hope its not too late. I dont think its ever too late, so good theres a whole lot that you can do in the backwoods. With no one watching just kidding but anyway on that note, thats gon na do it oh weve heard thats gon na. Do it for us today, dont take that and run with it, okay, its just a joke: uh yeah. We might just open it up. To that whole mean peoples, comment section and on that note, were gon na end the show. So i hope this helps you guys uh. Let me know if it does or not. I would love to hear feedback if other people have uh hey. You know what someone chimed in actually and talking about all the emails i have yet to respond to um.
There was a guy who said he had ag knowledge from california. Maybe this would be a good one for him, so yeah anyway, um uh. Thank you again for joining us. If you have a question, ask adroniu but thats gon na do it for us today.