Creating Ground Control Points


#21

Did I understand right that the ground control points are in place and assigned accurate coordinates already? So you don’t need equipment to survey them? If you do need to survey, the minimum equipment requirements in order to have high accuracy ground control points would be around $500 USD plus ground control points. (painted plywood squares)That would require post processing static gps data on a computer after the flight to calculate the location of the ground control points.

  • If you already have access to the location coordinates of the gcps you only need to identify them when you process the flight photos in dronedeploy and enter the coordinates there.

  • If you don’t have access to software for processing imagery with ground control points, just use the free version and export the image to QGIS.(free open source software) It has features for georeferencing pictures but it won’t help with accurate volumetrics. Also, I can’t guarantee just how the accuracy of the image will be affected by QGIS’s georeferencing process.

  • Here is some info in dronedeploy’s own words. Download the whitepaper for some good info on gcp points.

  • Dronedeploy GCPs


#22

Sorry for commenting on an old thread, but I don’t want to loiter the forum on here. I read all of this, and the way I understand it, when using an onboard RTK system on the drone, you don’t need to set GCP’s?


#23

You do still need GCP’s from my understanding. The RTK is on the drone itself and improves the accuracy of that piece of the map making. This means your map overall is more consistent in relationship to itself. I can see that you wouldn’t need as many GCP’s though. On 50 acres I currently use 10, but I could see only using 5 with RTK. Basically just enough to encompass the site and one in the middle like your would do with a traditional GPS survey system.


#24

If using ‘true’ RTK on the uas, no GCPs are needed to be added to the processing of the orthoimage for it to be accurate if the RTK system is working correctly. However, GCP’s are usually placed in order to verify or “truth” the orthoimage after processing it.


#25

Do you know of any RTK systems that actually geotag the photos with the RTK and not the standard module? I don’t think there are any DJI’s that currently do. The only other option is PPK and the images are still not directly tagged.


#26

No, I don’t know of a dji craft or aftermarket modification for RTK on a dji craft. Since RTK would need yet another com to a ground station or network, it really makes much more sense to use one of the precision gnss/ppk solutions available. While the images are not geo-tagged with locations from the gnss receiver in real time, the image geo-tags are replaced with the tags from the gnss receiver prior to the photogrammetry processing. So the result is the same, no GCPs needed for processing, just for verification.


#27

Have you tried this with DroneDeploy? Can you provide a map as an example?

@zach1 can you provide some clarity on this subject?


#28

I have not purchased a system. They are not cheap, a minimum of $6k unless you build one yourself. The ortho work I do does not currently justify the cost. Someday maybe. If you have the budget for this it would most certainly up your game, no doubt about it.

The technology is processor agnostic though. It shouldn’t matter if you process in DD, Pix4D, Agisoft, whatever. How it works is that each image is accurately geotagged in 3d space with centimeter accuracy rather than the many meter accuracy that is typical. So, the resulting ortho will be centimeter accurate without ground control because there is no distortion for the processor to correct.

You then use your surveyed GCPs to verify that everything worked as it should and can even certify the results if you are licensed.

I would note here that you need to calibrate the camera for this process too. The online processors, DD included I think, have built in correction. But it is not as precise as an actual calibration of the camera itself. And they don’t allow for input of camera calibration corrections. Pix4D and Agisoft have that ability. DD and the others would need to add this if they want to play in the same league. Maps Made Easy has told me that they are not interested at this time and suggest using one of the software solutions I mentioned for serious survey work. I suspect DD will have the same response.


#29

Were are using a $2k Phantom 4 Pro setup and regularly get RMSE corrections (of the camera) under 0.5 inches. The camera calibration input on Pix4D attempts to do the same thing, but is basically voided if you use GCP’s. If you use the Pix4D camera calibration input without using GCP’s you will see very little accuracy gain over DroneDeploy’s process without a manual camera calibration.
What is it that you are trying to do that requires a $6k setup?


#30

I think what Dave is getting at is that using a PPK solution would mean that GCPs essentially turn into checkpoints, and are not ground control points. Regardless, you need some sort of RTK or PPK unit to verify accuracy of the resulting data. For high accuracy results, you need to verify, and not just take your processing solution’s word for it.


#31

Can you explain how RTK or PPK verifies other than through the metadata? We extract the metadata from the photos as backup to the point cloud as well as using GCP’s and checkpoints through DroneDeploy’s GCP tagging enhancement. The raw point cloud has thus far provided all the verification we have needed. I have run Kespry and a Matrice 210RTK in demos against what we are getting and the deltas are negligible. Dave did mention certifying surveys so that may be the linchpin in the use case.


#32

@MichaelL The workflow you are doing either with DD or Pix4D is typical for photogrammetry using a “normal” p4p. Even using a DJI drone that is advertised as “RTK” is no different. DJI’s current use of the term is misleading and not accurate.

If you were to use a precision gnss system on your p4p, and you have your camera calibrated, you will achieve highly accurate results without the need for GCP in the photogrammetry process. I will say again how this is possible.

With a normal p4p, when it snaps a picture it records it’s (the p4p’s) location in the image meta. That location could be meters off from the 3d point it records to the image. So when you process, the resulting geotiff can be meters off. To bring that geotiff closer to accurate we add suryveyed GCPs to tell the processor where certain points on the ground are “really” located and the processor adjusts the geotiff to account for them.

With a p4p equipped with a tuned precision gnss, it records the location of where the image was taken in 3d space very accurately. Not meters off. When you are done flying the mission, you have the image set with the meta recorded by the p4p and you also have a log file from the precision gnss system. You use software that takes the images and the log from the gnss, and the log from a survey network station as close to your site as possible, that first applies the corrections from the local CORs network (this is called PPK) and then takes the images and replaces the location meta with the ppk’ed locations from the precision gnss.

So, now you have your image set and the location data has all be replaced. Now, when you process with DD or Pix4D, the camera locations are very precise and not meters off. The resulting geotiff will (should) be very precise without adding GCPs into the mix. To “prove” your results, you take your surveyed points that you used to call GCPs and use them as “Check Shots” What that means is that you don’t add them to the processing. After the processing, you select any of those points and it’s location on the geotiff should be very close to it’s surveyed values. That “proves” your geotiff is accurate.

This is still a simple explanation but hopefully it will be enough for you to get the big picture in order to be able to research if further. There is a big difference between using the gps on the p4p to write locations to images when compared with replacing those locations with precision gnss/ppk corrected locations.

DJI is ready to release an RTK P4P. We will have to see if it is a “real” RTK capable craft or not though. And I would bet if it is it will cost a bit more than the normal p4p. :slight_smile:


#33

@Dave, Thanks for that! I am well versed in survey/construction GPS so I agree with the majority of what you said, but the fact still remains that we were talking about two different things and we are both right. DroneDeploy is a DJI-only solution so as it remains for this use case RTK is positional RTK, not site referenced GNSS RTK. From the currently available links it doesn’t look like the P4P-RTK is going to be a referenced base-driven solution, but the positional RTK that is currently available with the M210. We can keep our fingers crossed though because even if it is a couple of $K more I would gladly pay it over setting and maintaining the 400+ GCP’s I do a year. :wink:


#34

Michael, I may be misinturpreting your reply so I’m sorry if I am. My point was that there are precision gnss/PPK options available ($$) for you to use your existing p4p and still process your set in DD or whatever processor you choose and NOT need to use GCPs other than for proving (as check points) your results.

If this system is working properly and you don’t need to prove it, you can forgo the surveyed points completely or at least need less of them to check your work.


#35

No, that completely makes sense. Can you provide a link to such an upgrade? I would be interested in pursuing it if it is what I think it is.


#36

Sure. Here is one.
http://www.airgon.com/loki.html


#37

Thanks. Is this what you use? They will be calling me Monday.