GCP vs RTK Drone Accuracy
ฝัง
- เผยแพร่เมื่อ 29 พ.ค. 2024
- In this video I look at the accuracy drone photogrammetry using GCPs and RTK and compare the two.
00:00 - Intro
00:34 - Field Methodology
01:24 - Connecting a to a Trimble R10 Base
02:16 - Verify RTK connection in the Pilot 2 app
02:45 - GCP color
05:26 - Defining the check points from export data
06:41 - GCP only results
08:15 - RTK flights
09:55 - Losing RTK connection mid flight
11:04 - Maximum errors observed
12:37 - Excluding control points from accuracy comparison
13:05 - Deeper dive into observed errors
14:49 - Extra checks
16:05 - Final observations - วิทยาศาสตร์และเทคโนโลยี
Thank you for this comparison!!! Awesome data
You know in school as a kid in the 1970’s and beyond there would be a picture and a name to introduce a subject. You my friend are one of those folks in history deserving of that recognition.
Strong thinking and excellent presentation. Keep going with your ideas and presentations.
In my late 50’s I am learning GIS, spectral and spacial imagery. SAR sounds amazing but I have not found a business use for my applications as of yet.
Thanks for puttimg a lot of maths together here. I am a Trimble user but a very recent adopter of the M3E RTK. Here in UK drone legislation is complex and confusing and it is unlikely I will ever do aerial work on paved urban surfaces but it is perfect for large areas of moorland and mountain side. I have just completed a 100Ha survey on such terrain with drone hooked up to VRS. VLOS regulations would otherwise dictate I would have needed to shift R10 base 7 times. While I had GCPs on exposed rock slabs, in the end I used them only as checkpoints along with other spot heights taken when marking GI boreholes by RTK. I made comparison against the extracted ground mesh which tries to stick to the underside of the PC. My checkpoints fitted well against this. Some were in long grass or heather. My gut feeling is that DJI terra does not like a GCP that sits higher than surrounding ground as the one I had in this situation was worst for level. I was completely happy with the ortho image which I deliver along with the DTM. The deliverable does not have the accuracy of a total station survey but the client doesn't need it for early design and volumetric. I have more faith in the RTK approach now than I did a week ago and your conclusions here gave me some comfort in that view. Thanks for your detailed breakdown.
Consider retaining image locations even when processing GCP-only data. The removal of image locations before processing may introduce unnecessary inaccuracies, as these locations could provide valuable information for data validation.
Also, maybe address how accuracy can be effectively reported to the client when using GCP-only data. Discuss the importance of checkpoints in validating results and ensuring data quality. This information is crucial for clients to understand the reliability of the output.
Yea that might help with alignment. It might also hurt accuracy when the software is determining camera locations by introducing much less accurate data, if you are talking about non RTK tagged imagery, if you're talking about RTK tagged imagery then that would infringe on the whole point of the video and skew results. I wanted to compare accuracy if you have an RTK enabled drone vs. not. The perk I wanted to highlight is that if you have an RTK drone you can opt for not having to set GCPs, check shots are of course always needed.
Absolutely, I believe I address check shots several times in the video. Providing spatial data like with without INDEPENDANT check shots is very poor practice in my opinion. GCPs or not you must always have ground truthing shots.
Thanks for the comment, I always enjoy trying to get discussion going in the comment section!
We kept having our pvc sheet targets stolen and so we had an aluminum target template cut for us. We started experimenting with different paints and found that Black and/or red chalk paint works the best. The chalk paint is extremely flat and we didn't get the reflective shine when at certain angles in high sun. It's a little more expensive and messy but works so much better.
Very interesting! I've never tried matte paint before but I would imagine it would have a similar effect. Thanks for sharing!
Great examples and breakdown. Thank you.
very informative. thanks for the video. I've been absorbing a lot of videos and I like this one a lot.
great video very helpful for some future work My company is looking to get into.
Excellent information, thank you!
Great video, excellent information.
U need to publish this, u already have the data, great video!!!!
Great Video! I hope you would not mind sharing what software you are using for processing your flights. I use 3D Survey and am having some issues utilizing the new Mav 3E RTK feature. Do you have the capability to use both the telemetry from the pics along with ground targets? My ultimate goal is the be able to scale, rotate and perform a datum adjustment and achieve this prior to producing the point cloud. I would really appreciate your feedback.
Hey Mark, I use Agisoft Metashape. It has been fantastic so far. Yes you can use picture metadata and GCPs, that's what I did in this video for some of the testing. With Agisoft you can always do a custom transformation when exporting as well.
@@The3rdDimensionSurveying Awesome! I appreciate your help
Did you by chance compare the built in error reporting from the mapping software to the technique you developed? I'd be interested in knowing if the built in system is at least close.
None that I saved that I can share now unfortunately. What the software reports will be overly optimistic. You can see at 0:59 mark the total elevation error for all GCPs for example was 0.007m. I'm guessing this may be the sum of the residuals for all GCPs in the vertical direction? I wish software companies were more transparent about what these numbers actually mean.
But if we can't set our GCPs to 7mm then how can we expect our final result to be that good after we add on many other sources of error?
IMO these numbers can be used to compare the quality of one data set to another or use to search for blunders but they are by no means a measure of absolute accuracy one should hand to their client.
Nice info
Good morning, would you be willing to help with getting the Mavic 3E to come in on the right coordinate system. We have been loving our 3E, but trying to setup the base on a known point and get it to come in Pix4D in the right place has been a bear. Thanks!
Is it shifted a few meters or is it out more then that?
Man, beautiful work explaining the corrections with GCPs and/or RTK.
About your marker colors, what you showed was mostly on paved surfaces (asphalt, concrete). What would you suggest directly over soil (brown dirt, or sandy soil) and grass?
Black and white mesh targets are always a great option in that situation. I also often use 3' wooden stakes and paint the tips and arrange 3 so they are all pointing at a control points spaced out evenly like a triangle.
Yes, I have some 20" black rubber mats and painted white tiles. I thought on Aerotas targets, maybe I'll try mixing blue/pink, orange/black@@The3rdDimensionSurveying Thanks
Hello, its a really intresting video, wenused our M3E for about 4 flights now,(ofc with the rtk module) and we used a goverment provided source. The Photogrametry sofwarte (reality capture) puts the model about 2 m higher than it acualy is… is that a problem you also run into?
Thank you for you intresting video!
Sounds like a Geoid issue, do you have the correct one selected selected? Remember the GNSS corrections you're receiving will be ellipsoid heights.
Thank you, is that something with RTCM 3.1 /3.2?
No, you need to apply the correct Geoid model to the EXIF data of the imagery so the elevations are referenced to the datum you're trying to compare to@@kohleholeHD
Just started messing with our new M3E yesterday and tried connecting it to our base station , found out that its a little more complicated than bluetooth or making sure its on the same radio frequency. I saw that you used the Trimble GUI to broadcast your Ntrip signal , is that accessible thru the M3E controller? I have TBC in my office with a license , however the crew only has a laptop and no TBC , we alse use Spectra Precision SP80s for RTK so I am not sure if its even compatible. I see there are NTRIP casters like Emlid and others we could sign up for , and we also have a subscription to Smartnet VRS I figure I could connect to , but I am not sure how that compares to using onsite control versus a VRS base miles away. Any thoughts on a direction we should head to be able to connect the M3E to our SP80 RTK? Just started messing with this late yesterday! Any input is appreciated!
Yes that was the controller screen you were seeing. Just punch in the IP address I put in and enter the username: admin and password: password
@@The3rdDimensionSurveying I will give it a try Monday 🤙
@@The3rdDimensionSurveying Question , are you using a cellular connection in your base station to transmit out the corrections? I was messing with it yesterday and was trying to use the internal Wi-fi connected to the phones hotspot to transmit out since we dont have a SIM card in the base at the moment , I basically went in circles trying to get it to connect...Thinking the base needs its own SIM...
@@butcherben12 No, I am not. I use the internal radio to broadcast/receive my positional corrections. The wifi (which does not broadcast corrections) should be fine to connect to the mavic controller (and of coarse the data collector). A sim card, like the internal radio, is used to receive positional corrections from a service that is broadcasting them from a known control point, the benefit of using a sim card is now instead of a radio connection to your control point you are using a cellular connection which obviously allows for much longer baselines (100 miles instead of 2 miles). Not to connect to antenna or GUI.
Can a Trimble R8 model 4 be used as a base just like you did with the R10? Looking at upgrading from my emlid Rs2+. Hard to believe that a 10 year old R8 would be an upgrade over the new RS2+ :)
As long as it you can connect to it via WiFi and the Web UI and it can Cast NTRIP, then it should be able to. Just make sure you use the correct settings on the base and the drone.
Anyway to get a copy of your findings like uploading it to Google sheets and making it public viewable or making a pdf version so we can print it off to show clients that have guesting. But also so I can look it over more in depth
I can keep that in mind for future videos for sure, I don't have any PDF available for this data unfortunately .
Would PPK processed flights hold the same accuracy results as RTK
PPK yields better accuracy, it just takes more effort and time. It also has the benefit that you don’t need to maintain radio communication between each receiver.
I see your software you have on your web site. Do you sell subscriptions?
Not at this time, I’m still in the development stage. Adding new tools and features ect
So, who are your clients (industries, I mean)? Construction, I'm sure. Maybe land developers? I would think that most modern surveyors would invest in their own equipment (correct me if I'm wrong). Just trying to understand who the market is before I invest $15K in equipment and spend months training to start up a business. This type of work is right up my alley, though!
Well I am a Land Surveyor, so I serve all the same industries a typical surveyor would. Most "modern" surveyors do typically have a drone program and they are in fact my target audience for these videos. I really couldn't recommend someone without a deep understanding a survey measurements and the errors associated with them, scale factors and coordinates systems, and how to properly QC your data try to "survey or map" with a drone. On a typical job I'm using a lot more than $15k worth of equipment and months of training may be a bit off. I have a degree in Geomatics and 15 years of experience and am learning new things every day. A lot on the TH-camrs trying to push drone surveying as an easy way to make money are doing a disservice to the profession and their viewers in my opinion. I've been on more than one job where an inexperienced surveyor didn't fully understand grid to ground scale factors (one example of many) and because of their mistakes associated with that were being sued for a significant amount of money. Surveying usually comes with quite a bit of liability.
@@The3rdDimensionSurveying Yikes, yeah, I don’t want to put myself in that position and I don’t want to do survey work. Thanks very much for taking the time to reply. I’m retired from a career in Computer Science and have been a licensed drone pilot for 5 years in real estate photography. I’m investigating photogrammetry as a possible area to extend into, using platforms such as Drone Deploy, but it looks like I have gotten slightly off course. I found your video while searching for RTK vs GCP topics. Sorry if you feel insulted by my honest (if somewhat ignorant) question.
@@mediamannaman Not at all insulted, with the amount of people pushing the idea of using it to make an easy buck on TH-cam there are an increasing amount of people jumping into the industry, especially with the amazing quality and relatively low cost of drones coming out. I always wondered if there could be a market for using photogrammetry in the real estate market to provide clients with a point cloud or textured mesh where they could manipulate the property in 3D. They could even pull *approximate* measurements (just throw a disclaimer in with the data) and maybe even overlay it in UE5 with google earth data or you could place design building on vacant lots. Maybe an idea for a future video...
@@The3rdDimensionSurveying Yes, exactly. Thanks. But also expanding to something like early stages of large construction projects, where calculating the volume of soil and base material in large piles would help manage the project. I’m just trying to learn what the possibilities are and I bumped into a side rail with your video, but am still subscribing for what I can learn that would apply to these other use cases for a Mavic 3 Enterprise with an RTK module (and/or GCP’s).
Vernon!
Well... Coldstream, but good eye ;)
@@The3rdDimensionSurveying im am trying to better understand Base stations etc... Is the RTK it self on the aircraft not enough?
No, it’s not. You need to get corrections from another gps receiver that is setup on a known control point and broadcasting those corrections to the drone via a radio link. You can setup this base station yourself or subscribe to a service that provides these corrections. I use my own base station.
@@The3rdDimensionSurveying I just saw the cost on the Trimble Base Station... yikes !!! I just got a Mavic 3M with the RTK. Comes with the package.. Not too sure if i will need it or not. Trying to find out as much info as i can.
Not at this time, I’m still in the development stage. Adding new tools and features ect