At first, I did a grid scan with a DJI Mini 2 to see the potential of 3d mapping with a drone. I have been really impressed with the output. (with Drone Deploy)
So I decided to upgrade my drone with a Mavic 3E RTK.
At first, I was going to buy a DJI D-RTK2 base station, but from what I can read online, RTK isn’t necessary the best solution for my project.
I have read a lot online but I have a hard time figuring out exactly what I need. Reach RS2+?
My objective, see picture below, is to scan in 3d my two plants to get a 3d model that I can work with for:
3d crane lift plan
Structural/mechanical design of an expansion plan for my plant
I will do several drone flights. (grid scan from top, vertical flight for buildings, piping, stair, catwalk, equipment, etc)
I would like to get the more simple and fast setup possible to get a good accuracy. If possible, without GCP but I’m open to it. At this moment, I don’t have known coordinates to establish my base station.
Actually to achieve real world coordinates, you’ll either need a base onsite or use an RTN network to tie your imagery to the National Spatial Reference System if you are in the USA. You’ll also need checkpoints on the ground (GCP’S) in order to verify your imagery accuracy.
Another way to verify the imagery if you have availability of existing LIDAR data is to compare your surfaces. I usually do this on each flown site even with GCP’s. Accuracy is usually 3-6cm comparing hard points (street intersections, driveways, etc ) of the existing LIDAR. We’re fortunate to have LIDAR flown by our state with the latest data done in 2020.
I use GCP’s in our Matrice 350 with the LIDAR sensor and also with imagery with our Autel Evo II. The accuracy we usually obtain is 3-6cm with verification to the GCP’s. You can fly the site with RTN without GCP’s but you’ll never know the true accuracy of your imagery and surface models.
If you need your model to align at centimeter level with a real world coordinate system, then see Bryan’s answer above. If your goal is to have an accurate digital model of your site, but it doesn’t have to align closely with a CRS, then you can likely get away without any additional gnss receiver, especially if you have CORS stations near your site that you can obtain observation logs from in order to post process your camera locations.
I don’t know about Drone Deploy, but in Metashape, you can confirm the accuracy of your model with the use of scale bars. Which are basically known, measured distances between objects in the model.
You’ve got an RTK drone but without a source of corrections - either CORS or local base - you won’t benefit from the available positional accuracy.
I went through this, thinking I could just rely on CORS. Ultimately you need both a source of corrections and GCPs if you are to produce accurate data. I’m very pleased that I finally accepted that and bought a Reach.
Your statement is correct in the case where the data needs to align with a CRS. “Relative” accuracy, especially for a relatively small area, is not so dependent on gnss correction. But it doesn’t hurt.
You would fly the drone with Reach RS2+ and RTK? There is a lot of massive steel structure, I’m concerned that it could make connection between base ans drone unreliable.
I think you should elicit his views as probably more useful than mine.
For me, I’ve simply found that running GCPS delivers tighter relationships in the orthophoto and everything is more accurately related to its neighbours.
I perhaps didn’t say the other piece so well: my experience is that without RTK corrections to the drone in flight the positional accuracy is not necessarily any better than a consumer drone…or your phone. If you are flying around structures then it’s nice if the drone goes to repeatable places at decent accuracy.
As for the PPK questions, Dave is probably the man to answer that.
Got it. Since you will be flying to the same site, and absolute accuracy is not a strict requirement, you would only need to have your models as accurate as possible relative to each other.
Here’s what I can suggest:
Set up a permanent benchmark for your base. This will be where you will set up your base every time you fly the drone. You can do this by setting your Reach receiver over a point, recording logs for a few hours, and uploading the RINEX file to a PPP service. Otherwise, you can also use Static Processing in Emlid Studio by downloading the RINEX from the CORS.
Set up “permanent” GCP/CPs around the site. These will be the points where you will place the markers. Since you’ve mentioned only using 1 RS2+, you can establish GPCs using the Stop and Go workflow.
I forgot to mention that even if you only need relative accuracy, it’s better to have the base coordinates and the GCPs in the same coordinate system. Otherwise, your model will likely be shifted relative to the GCPs.
For point 1, I suggest setting up the base (RS2+) over a permanent, known point every time you fly the drone for consistency. Since you will only have 1 GNSS receiver, the only way to obtain its precise coordinates is with PPP or Static Processing.
For point 2, to avoid collecting GCPs every flight, I suggest setting up permanent GCPs on your site using the Stop and Go workflow.
This really depends on what the project CRS is but is usually not that big of a deal. It’s kind of the whole point of GCP’s. If it is close to a natural projection you can single-average a point and the GCP’s will take care of the rest. If it is an arbitrary (local) coordinate system the only problem will be being able to initially locate the targets, though if you are using a system that allows you to manually choose images it will localize and after two or three tags will find the rest. You can also fake in a coordinate manually that is close to what the project is using.
As @ruth.bongon also mentioned absolute accuracy is not necessary if you are just doing progress and comparing flight to flight. It is not really that hard to go the other direction and get Design data back to the drone’s origin if you have the right tools but it does cause a break in the natural data sharing between Design and drone ops. Still though drone operations can be completely segmented from the site data and be very valuable.
