Hi everyone!

I'm in the market for an RTK-drone that will be mainly used in the roofing sector. Other than simply inspecting the roof, I will take pictures and use a photogrammetry software to create a 3D model and thus have access to all important dimensions. While the software is easy to use and I've seen how it works after using pictures taken with my Mini 2, I'm currently considering the Phantom 4 RTK drone and the Mavic 2 Enterprise Advanced (EA) with the optional RTK module.

However, I'm not sure which one of these is the best for my use case. The thermal camera on the Mavic is a very nice addition, but I heard someone on YouTube saying that the Mavic 2 EA (even with the RTK module) isn't as precise as the Phantom 4 RTK. Note that I'm not planning to use a base station, so I can understand why the Mavic 2 has a disadvantage on paper. But what else could impact the accuracy of the Mavic drone? I was thinking that it might be less stable, or that its camera is built differently to that of the Phantom.

I would highly appreciate an answer from those of you that have hands-on experience with either of these drones. Thank-you!

You should probably educate yourself on what RTK/PPK GNSS is and how it works so you can make an intelligent decision on what gear is right for you. Your statements and questions suggest that RTK is a bit of a black box to you. You say "I'm not planning to use a base station", but do you realize that without a source of corrections (in real time or after the fact), RTK/PPK GNSS is not much more precise or accurate than a "normal" GNSS? The corrections don't have to come from an on-site base station, but they need to come from somewhere (NTRIP, PPK, basestation).

Do you need cm accuracy? Do you understand how to use GCPs to improve and/or assess the accuracy of orthomosaics and surface models created using photogrammetry?

Do you really need thermal?

In any event, the M2EA is not a good drone for mapping, because is lacks the TimeSync system found on the P4R, M300RTK, and other Enterprise level RTK drones made by DJI.
https://www.aerotas.com/blog/mavic-2-enterprise-advanced.

If you need high precision and accuracy, you need a P4R, and some way to provide corrections to it or use PPK. If you must have thermal, try the M30T (much more expensive). But before you spend a dime, make sure you understand what you're getting into & how this stuff works.

Thanks a lot for your reply!

I have done a fair bit of research beforehand and had conversations with people that used photogrammetry software. The discussions I had with them revelead that I don't need to bother with GCP's or base-stations as the RTK module can deliver 2-3 centimeters of relative accuracy (which I believe is also known as precision). I don't care about absolute accuracy at all, but perhaps they haven't told me that I need to perform PPK anyway. It's something I will have to research further because as you said, RTK is more like a black box to me at the minute. (I consider it a better GPS that "just works" out of the box, which is probably not the case.)

That being said, I now understand how the cameras on these drones differ and why the Phantom 4 RTK has survey-grade accuracy. Before you replied I learned that the mechanical shutter plays an important role and it differentiates the P4R from the M2EA.

Lastly, the thermal camera would be awesome as one can easily detect roof leaks. However, given that neither the M2EA nor the M30T have survey-grade accuracy, the only option that I have is the P4R (other than the M300RTK with various payloads, but that's just overkill).

Good for you! Sorry if I sounded lecture-y, there's no small number of folks on this forum new to differential GNSS and geodesy and all the things that come along with trying to do survey-grade (or at least high precision) mapping, and sometimes I go into "do your homework!" mode.

RTK (and PPK) are methods of differentially correcting GNSS measurements. With a proper satellite constellation, the poor accuracy of standalone GNSS is mostly due to atmospheric noise, resulting in generally ~5m uncertainty. Differential correction relies on a 2nd GNSS unit fairly nearby that has been precisely surveyed in, and "knows" it's true location; it compares the position it calculates from the satellites to it's known position, and generates a correction (the "difference" in differential). This is applied to the position of the rover GNSS (drone) in either real time (RTK) or in post-processing (PPK). For RTK, the corrections can be provided via radio from an on-site base station, or over the internet from a nearby NTRIP base. For PPK, data are downloaded frop a Continually Operating Reference Station (CORS) and applied after the fact. But without some source of corrections, you are not going to achieve 2-3 cm precision OR accuracy.

https://docs.emlid.com/reachrs/rtk-quickstart/rtk-introduction/

https://novatel.com/an-introduct ... -time-kinematic-rtk

The rolling vs. global shutter concern you mentioned is the least of the problems with the M2EA wrt mapping. As I mentioned, the real issue on the M2EA is the lack of proper time sync b/w position, attitude, and gimbal orientation. This precludes high-precision image positioning. You know where the aircraft is to within a few cms, but that precision does not extend to the images captured by it's camera.

I agree the P4R is probably the tool for your application, just be aware that there is a little more to realizing its potential than just opening the box and flying it. Regarding not needing GCPs, the only way you will be able to truly tell what you relative and absolute accuracy is is to have some checkpoints or scale bars or something to use as a reference. You don't need them so much to provide accuracy as to assess it (but you can do just a a few rather than a whole array).

Good luck!

You don't have to be sorry about anything! In fact, you've been a lot more helpful than the guys trying to sell me the drone and/or software: neither of them mentioned that I need to perform the corrections, so I am very thankful for all the information you shared with me. I knew right from the start that my knowledge wasn't quite complete and therefore I didn't take your post personally ;).

With regard to the checkpoints, I've already scaled my 3D models based on a given length and I've been quite impressed with the precision I obtained (it's almost spot on in most cases!). At this point I feel like the combination of Mini 2 drone + 3D photogrammetry software + scaling works very well, but I will have to test this conjecture on more complicated roofs (especially high-pitched roofs).

In the meantime, I will keep reading upon RTK/PPK using the links you provided. Thanks again!

fans4f10f52a Posted at 6-5 15:07
You don't have to be sorry about anything! In fact, you've been a lot more helpful than the guys trying to sell me the drone and/or software: neither of them mentioned that I need to perform the corrections, so I am very thankful for all the information you shared with me. I knew right from the start that my knowledge wasn't quite complete and therefore I didn't take your post personally ;).

With regard to the checkpoints, I've already scaled my 3D models based on a given length and I've been quite impressed with the precision I obtained (it's almost spot on in most cases!). At this point I feel like the combination of Mini 2 drone + 3D photogrammetry software + scaling works very well, but I will have to test this conjecture on more complicated roofs (especially high-pitched roofs).

You're very welcome. The easiest way to do RTK without setting up your own base station is to get an account with a free NTRIP provider such as CRTN (if you are in California) or something similar in your area if available. Then all you need is an internet connection and you'll get realtime corrections duirng your flight, so that all your image position data will be high-precision.
PPK is also a viable option for obtaining high-precision without setting up your own basestation, and in some cases can be more robust than RTK (esp. where internet connection is spotty if using NTRIP). But PPK takes a few extra steps (hence the name) and does not yield high-precicion in realtime.

Have fun!