Hi,
while I saw this thread:
https://diydrones.com/forum/topics/using-a-parrot-disco-for-aerial-mapping
I am wondering if there are some updates on this.
I created a new thread as the old on is already long (so is my script below ;-) I'd appreciate if someone could help with this. Maybe this could turn into a beginners cookbook for this type of usage of the Disco?
Brief background info:
I bought a Mavic Air a while ago and when offering photographies over here (Germany) to my clients got approached to fly a mapping mission over 500acres of forest.
So I started looking into fixed wing drones. The possible solution needs to be low budget, as there is only this client for now (even though I am surrounded by farmland, I might find more..) and I have not enough funds to go expensive right now.
Second hand Parrot Disco consumer models are very cheap now, so this will be the airframe of choice.
I am not at all a programmer but, as the consumer version of the Disco has no proper mission planner I am now looking into Ardupilot.
QUESTIONS:
As I am new to this Ardupilot & the programming thing, I need some not too technically worded guidance. I run on Android & Windows 10.
A) Mission planner, which one?
I noticed there are several mission planner apps around. Which one is the best suited to someone not firm in programming and really needing a proper GUI? I'll fly primarely fixed wing grid type missions and want to fly fully automatic. I really lack the know how at this stage to understand the details.
So far I found:
http://qgroundcontrol.com/
http://ardupilot.org/planner/
http://ardupilot.org/planner2/
http://qgroundcontrol.com/
Some comments of what I found so far:
-QGroundControl has:
Most intuitive user interface, but I cannot find a way to set it to fixed wing.
Custom camera spec can be easily added.
90Deg offset of pattern greyed out.
-Ardupilot
Has a lot of options to customize a flight plan.
Lots of different flight plan patterns
Good user interface.
Custom camera spec can be easily added.
-Ardupilot 2 (APM)
I could not find a way to create custom survey patterns automatically.
Feels unfinished. Possibly still in development.
Any other & better suited ones?
B) Camera, which one?
I have looked into cameras, my clients will most likely need:
Orthomosaic RGB, NDVI and, in case of forest, a 3D model.
One question upfront, can one receive the front camera when flying with Ardupilot and an after market remote control?
Unfortunately the best lower budget multispectrumsensors are still quite expensive (Parrot Sequoia & Micasense Red Edge) so thats not a low budget solution.
Mapir does some interesting cameras too, but their Survey 3 model does not seem to save IMU roll angle in the images. On a fixed wing without gimbal that seems an issue.
Is it somehow possible to add that roll data to the images via data from the CHUCK? Automatically?
If yes I am thinking about putting these two onboard (possible swapping them and do two separate flights).
https://www.mapir.camera/collections/survey3/products/survey3n-camera-orange-cyan-nir-ocn-ndvi
&
https://www.mapir.camera/collections/survey3/products/survey3n-camera-visible-light-rgb
Did someone install a downward facing cam with an added gimbal to the Disco airframe?
Alternatively which GoPro (or something completly different?) is best suited? Do they have IMU roll data?
Does the increased internal stabilisation of the Hero7 make it much more desirable to use than the Hero 5 or 6?
Does one need to replace the lens on the GoPro to take out the fisheye?
Does it matter if the NDVI is taken with a different lens than the RGB (like flying RGB with a GoPro & NDVI with a Mapir)?
C) Batteries?
Obviously its a compromise of extra weight vs power? S
till for the 500acres I need a long flight time (or two flights).
I am leaning towards this one as a compromise, as the drone has to carry the extra cam & receiver & tracker as well.
https://hobbyking.com/en_us/multistar-high-capacity-4000mah-3s-12c-multi-rotor-lipo-pack-w-xt60.html?___store=en_us
Any better suggestions?
D) Remotes?
As I understand I need an after market remote to fly with Ardupilot.
Something like the FrSky Taranis X9D.
What receiver model do I need exactly?
There is no other simple way to upload a mission plan to the aircraft otherwise?
Using the Skycontroller (which comes with the drone) to start the craft and than have it fly fully auto is not possible?
E) Post processing?
I'll lean towards Pix4D or WebODM.
The missing IMU roll angle of the Mapir camera seems an issue. How about the Gopro's?
Any thoughts on this?
Thanks so much for any input, suggestions & help you can give to a Newbie.
Sorry about the long post,
Fran
Replies
Btw, I just found out that terrain following is possible too.
http://ardupilot.org/copter/docs/common-terrain-following.html
Hi Darrell & Andreas,
more great info, thank you.
I had a meeting today with my client. We did not sign something for now, but I think he is fairly interested and understands the advantages and especially the need for true NDVI and not VARI or fake NDVI. He has to digest the pricing though ;-) . Still running the smaller airframe of the Disco makes my prices more compatible.
Guess in the end it might be the Disco AG with their standard mission planner.
We'll see.
He's looking into expanding the contract to his neighbour as well. Might invest a little more than, we are talking 1050Ha (=2600Acre) than. Luckily he understood that I can not invest without a firm commitment.
Might take a while for him to commit though ;-)
Still, I will study the Arduplane stuff further. Sounds interesting.
I looked into the different mission planners mentioned in my first post.
http://ardupilot.org/planner/ & http://qgroundcontrol.com/ are my favorites. The preconfigured flight patterns are great. I only wish I could display them as 3D with Google Maps 3D underneath. I learned today that the forest in quesion is along some hills. So I may need to built terrain tracing flightplans. Any way to do that? How does one georeference them than in Pix4D? Or should I just fly on one level and accept the differences in flight height?
Andreas, what you write about the CHUCK is also worth thinking about. The reduced weight could give room to a larger battery which gives a longer flightime. Your comment on weight centers is well noted, my original background is in yacht design and to some degree the same principles of maintaing weight centres applies.
Still, I would want to keep the frontal camera as well. It gives the client a chance to also look over my shoulder while flying. Psychological I consider this helpful for marketing reasons. Further it might help to see which tree you hit ;-)
Here: https://discuss.ardupilot.org/t/pushing-a-parrot-disco-to-the-limit... at the very bottom is a posting from 21 October 2018 by AlexP describing how the forward camera can be kept operational under Arduplane.
When looking at Dustin Dunhills long flights in Hawaii, I get the impression that a larger than standart battery and a second downward facing camera poses nor real issues.
Here: https://diydrones.com/forum/topics/using-a-parrot-disco-for-aerial-... is also a thread about how to make booting into Ardupilot can be made persistent.
Thanks again,
Fran
Quick reply because it's been a while and can't remember the details but Disco can run Arduplane very easily by following the instructions on the Arduplane docs. I remember there were a few bits where I had to google things (I remember I followed a video tutorial that I got on youtube, it *might* have been this one) but perhaps docs have been updated since then. Basically, it was a case of downloading Arduplane on your computer, logging into the Disco autopilot and copying two files into a particular folder. Then, when you boot, you do something with the button presses (can't remember exactly) and it boots into Arduplane, from which point on you can simply use Mission Planner or whatever Tower or whatever GCS you feel like.
Now, in terms of payload and mods, the Disco autopilot (CHUCK) is absolutely gigantic and the telemetry range (wifi) is terrible if you don't use the dedicated Disco receiver, which you can't if you are running Arduplane (unless this has been fixed, I don't know). If you are mapping, you will need to disable the GCS failsafe or it won't go anywhere but over your head.
Due to the enormous size and weight of the Disco autopilot, I would possibly invest in a much smaller one like the Pixracer (buy some appropriate radios for it for telemetry depending on your location) and rip out the cam from the nose and the wifi antennas from the bottom. You will save a lot of space and weight this way. Take care to note the all up weight and CoG location before you start ripping things out and then make sure you get it right for the new autopilot and whatever payload and battery you end up with.
The only issue with separate flights for NDVI and RGB is the time involved. The size area you are considering will take some time to cover so doing it twice would be quite a task.
It's been some time since I've used a GoPro and the newer models are much different from what I used in the past. If you have a small quadcopter I'd put a GoPro on it and do some test flights. That way you can get used to whatever mapping software you decide to use and how to process the photos etc with the least amount of expense. I think it would be time well spent.
If you are using a DIY quad then give Mission Planner a go. It has fantastic mapping routines built in. I use it for fixed wing and love it. If you like it, consider a donation to the author if you can. If you are using a commercial quadcopter, then you'll find the the software you use to stitch your maps will probably have a free app for mapping. Pix4D, DroneDeploy etc. My favorite, due to their pricing model and features, is MapsMadeEasy MapPilot. All of those vendors also have very helpful websites full of great info on how to do mapping.
Great info again Darrell. Many thanks!
The reason why I was considering the GoPro is that I also have other usage for it, it has great image stabilisation, is light weight, has inbuilt GPS & motion sensors, and is reasonably priced. Last not least, its fairly indestructible in case of a crash. At least on the last one the Sony would be a risque investment.
Very good info on Micasense vs, Sequoia. I was told by a guy flying with both that they usually get better results from the Micasense when it comes to resolution. I believe you though that the Sequoia is probably the better choice for me. Do you think its an issue to fly RGB and NDVI in two missions? Or do I need to fly both at once with two cameras on board triggered by the same signal?
As I only have the samll quadrocopter for now and hope to get also into more field stuff down the line, I feel investing into a fixed wing system is better for me. Usually these are large areas.
Thank you very much for all that great info. It really helps!!
Best regards,
Fran
A GoPro is a poor choice for RGB mapping. It's just not what they were designed for. The camera of choice for mapping is/was a Sony A6300. They are light, cost effective and simple to interface to a Pixhawk. You don't need to worry about roll angles as the stitching software will compensate for that. You don't even need to put them on a gimbal.
A MicaSense RedEdge really has no advantages over the Sequoia. They sell it as having greater resolution but the sensors are exactly the same. The RedEdge has a slightly longer lens so you get higher resolution at an equivalent height but you have to take more photos. You could do the same exact thing just by flying a bit lower with the Sequoia. The RedEdge does have a blue channel that the Sequoia doesn't have that you can use to create RGB maps but you'll find that the RGB from either camera is fairly useless. If you need NDVI and RGB you will find most people fly an additional camera to capture the RGB. The RedEdge M doesn't even have a GPS, you have to attach your own. But the newest model is the RedEdge MX which has a sunshine sensor that includes a GPS. The Sequoia sunshine sensor has always had a compass and GPS in it.
The Sequoia does have a nasty habit of the USB connectors breaking off the motherboard. I have two in for repair now due to this. If you buy one be very careful with the cables and make sure you never try to put a normal USB connection in the camera's connection to the sunshine sensor. So pay close attention to what cable you are plugging in to it so as not to cause damage.
As an aside, the only Parrot Disco I've seen actually had the choice to load Ardupilot firmware as an option. I don't know if they still do that or not. But in your case the Ag version probably has mapping software included so it's probably best to stick to that. Also, for the size of area you mentioned, you really could do your project with either fixed wing or a multi rotor.
Thanks Darrell,
what you say about sensors correllates more and more what I found when researching the matter further yesterday.
I even found a number of quotes pointing more towards the Micasense than the Sequoia.
So, if the client really wants NDVI in the end it looks like I have to bite the bullet and get one of them. VARI is just not (yet) a good replacement for true NDVI.
Weightwise they can both fly in the Disco Airframe.
Looking into prices I found so far are:
Parrot Disco-Pro Ag (includes the Sequoia) 4200€ ex VAT
Sequoia Sensor standalone 2990€ ex VAT
Micasense Rededge standalone 5041€ ex VAT
Parrot Disco Consumer Airframe with all equipment to run on Arduplane about 800€ ex VAT
So basically I could choose a complete (Sequoia based) solution for 4200€ or a makeshift solution with Arduplane for 3790€. The choice is obvious, in this case the commercial package is better, because it will work immediately.
If my client now would decide that he would only want RGB in the end, its a different subject.
RGB gives already a number of useful results to him:
-Area measures of new planted areas
-Storm damage asessment
-Treenumbers
-Treeheights
In this case I would get the consumer setup with Arduplane and a GoPro6 or GoPro7. A Sequoia or Micasesnse could be added further down the line than, when more clients show up.
Now here is two questions for the RGB scenario:
1)
Is the roll angle IMU stored in the images of GoPro 5/6/7?
Or is the superior image stabilisation of the GoPro 6/7 & the 80% overlap taking care of this enough.
2))
Is the fisheye charakter of the GoPro an issue when using it for drone mapping?
Does anyone have further input on this?
Thanks,
Fran
Ag mapping is a fairly advanced use case for a drone. It doesn't sound to me like a DIY solution is what you need. You will be spending a lot of time and money figuring it all out. You might be better off with a total commercial solution such as the Parrot Disco-Pro Ag. You will have to do multiple flights but it should give you proper results. I really don't think there is such a thing as a 'budget' solution for this application.
In regards to cameras, I'm afraid the Sequoia IS your budget solution. I believe they aren't selling them anymore so I'm not sure what will replace them. All the low end inexpensive single sensor solutions out there will only give you pseudo NDVI results.
In regards to your other questions it would take months to explain all your options. As I said, this is an advanced application of drones and it's simply not something you can just throw together. It takes a fair amount of experience and IMHO there just aren't any shortcuts.