RTKLIB with P4RTK

[amicron]

Does anyone know how to post process Phantom 4 RTK observations in RTKLIB? (And if you are going to mention Aerotas, I did every step exactly like they stated and it will not work. I contacted them and they say those instructions were a “public service” and they won’t address any problems with it or help me. They “don’t have time for non-paying customers.”)

I’ve tried it with my base Rinex files and CORS. It appears to process but the results are blank with only the headers. I can’t figure it out. I have three flights and none of them will process.

I hope someone has some better instructions or can help me troubleshoot the problem.

Thank you!

 

[doktorinjh]

Take a look at this thread and you can probably reach out to some of the guys there. It looks like there are three additional methods for you to try, other than the Aerotas workflow. If you're a reddit user, take a look around that sub, there's a lot of good information and knowledgeable people.

 

[amicron]

Yeah, I saw that but didn’t help and I’ve never used reddit. Thought I would try here first.

Thank you for the response.

 

[doktorinjh]

Yeah, I saw that but didn’t help and I’ve never used reddit. Thought I would try here first.

Thank you for the response.

So the Emlid workflow doesn't work, either?

 

[amicron]

ok, so I finally was able to get some results but I they are all over the place. I can only get it to run in Forward or Backward, not Combined. That was great until I ran it on each base we had setup (2 total, 1 Trimble R8, 1 Epoch 50) and with nearby CORS. Im getting 3 different results in forward and 3 different results in backward. So 6 different coordinates and elevations. Still nothing in combined. From what I can tell its due to lack of fixed and too much float.

Any ideas? Any suggestions? Please?

Thanks!

 

[PeteM]

I'm currently conducting numerous experiments with Phantom 4 (the simple one, not the expensive RTK one) to which I've just added a cheap Emlid M + module. I have developed a workflow, not very simple, the final goal of which are geotagged photos that are correctly ready to be inserted in the PhotoScan. The workflow I use is PPK and not RTK.

I got good results.

But I believe that the main problem with all these technologies that promise to eliminate GCPs is the impossibility of verifying the correctness of the topographic survey after processing. The only control method are a couple of GCPs.

But if I still must measure 2 or 3 GCPs then it changes very little to measure 6 or 7 instead.

 

[amicron]

I have this working now in case anyone needs help. Getting great results. I was doing several things wrong.

And regarding the last post, I will always have 10-15 check shots. If you want to claim a certain accuracy, check shots are the only way. You have to put in the work if want to prove your data is accurate. Check out ASPRS if you want to learn more about accuracies.

I like PPK because I’m more confident in the data and it allows me to put some constraints on the black box algorithms in the software.

 

[doktorinjh]

Getting great results

Great to hear! I'm glad that there are more RTK/PPK solutions coming online, but I think that the hardware is out in front of the software for the data crunching (looking at you emlid and Phantom RTK). We've been very happy with our Loki and eBee PPK systems, mostly because they have a very consistent and (nearly) foolproof post-processing workflow. Like you, we still collect plenty of checkshots and operate in PPK.

I talked to one vendor that tried to claim that RTK was the only way to go and that it was far superior to PPK. Their VTOL only had RTK, so I think they were very defensive over their methods. One of the biggest advantages to PPK is that we can set up a base on the side of a highway or a couple of miles away and then go find a safe place to launch and land. With RTK, you have to be tied to that base station, which isn't always convenient or safe...

 

[PeteM]

we still collect plenty of checkshots and operate in PPK.

When you talk about "checkshots", you mean "checkpoints" ?

 

[amicron]

With RTK, you have to be tied to that base station, which isn't always convenient or safe...

Or tied to an internet connection. More freedom with PPK.

I attended the ASPRS conference in Denver and the consensus there was PPK is best with RTK a close second. True DG with high end IMU is actually best but that’s a lot more expensive.

 

[amicron]

When you talk about "checkshots", you mean "checkpoints" ?

Those are the same thing to me... a photo identifiable point with known coordinates. And check points turn into GCPs if you hold them during processing.

 

[PeteM]

So you dont use any CGPs at all. Just many checkpoints to verify the final result.

 

[amicron]

I set flight panels and tie features and use them all as check points. Then if I need to hold some as GCPs it’s easy to do.

I’m a land surveyor and still fairly new to photogrammetry and drones so I’m definitely not an expert in this.

 

[doktorinjh]

So you dont use any CGPs at all. Just many checkpoints to verify the final result.

If we're getting good results with the PPK-only processing, then we'll use the our GCPs as verification/checkshots/checkpoints for our accuracy assessments. Like @amicron mentioned, if something goes bust, we can use maybe half of our checks as GCPs and salvage the survey. This happened a lot as we were figuring out the best processing methods, but we don't have to rely on those quite as much. If we're mapping, we tend to use a PPK workflow and set out ~10 targets and maybe pick up some additional photo points for vertical checks. That gives us plenty of data to create a reliable model with lots of verification points.

We will always collect more ground data than we need, but we're producing data for engineering clients, so it needs to be right. We see RTK/PPK as an additional layer of redundancy, but I trust it quite a bit. There probably won't be a case where we only use RTK/PPK without verification, but it's a great tool to have.

 

[amicron]

Nice. I feel like I’m on the right track. Can’t wait to have a workflow I’m confident in. I’m getting close.

 

[PeteM]

If we're mapping, we tend to use a PPK workflow and set out ~10 targets

Sorry but I still don't understand where's the advantage of investing money in the PPK stuff ?

Example.
In the traditional method, you place 10 GCPs: 7 to process images and 3 to check the final product.
In the PPK method you place 5-6 checkpoints to check the final product.
Savings is only a couple of points to be measured on the ground. I don't think it's worth it.

The money I should spend on RTK/PPK drones I spend in a system that allows me to quickly measure ground points (which is independent, so I can also use it on other occasions when the drone is not applicable).

 

[amicron]

From a land surveyor perspective, there is a confidence in accuracy and repeatability that must be achieved due to the liability associated with the data once it’s signed off and handed over. Land surveyors notoriously never trust the equipment 100%, are obsessed with reducing opportunities for error, and never want to make two trips to a site. We constantly do checks to verify or procedures to cancel out error - checking backsights, tie in existing control, two shots at 180 for GPS control, run levels, turn rounds, and so on. Seven control and three checks with a highly inaccurate standard P4P trusting the software to take all this inaccurate data and make it accurate does not instill the same confidence as obtaining accurate locations for original distorted pictures, having a camera calibration and having 15 well placed check shots.

Yes, a standard P4P can do the job and do it well but there are a lot of unknowns happening in the software. But for $3-6000 more (depending where you buy it) which is cheap for survey equipment, you can be considerably more confident in the data you collect and drawing you produce.

That’s actually the problem I see with this equipment and software being so affordable. And DroneDeploy and Pix4D and Aeropoints telling you how easy it is to create “survey-grade” maps. It’s not easy. You have to put in the work, learn how the technology works, test and verify the equipment and be confident you are getting accurate data to obtain high quality results every time.

 

[PeteM]

I understand your point of view but I don't agree with it.

If millimetric precisions, in each single point, are needed, the drone is not the right choice at all.

A distribution of GCPs covering the terraing in surface and height is sufficient to process the model correctly.
Some checkpoints for verification are sufficient too.

 

[doktorinjh]

millimetric precisions

We know we're not getting this level of detail, everyone knows that, but the more control we have over our error, the more confidence we have in the survey.

As previously mentioned, there are advantages to a PPK survey such as land access issues, safety issues, and irregular shaped areas where a standard GCP layout might not be possible. We fly lots of roadways. We can't always place GCPs outside of the ROW and only placing them in the ROW would not provide for good control. Or, things can happen to your GCPs while you're flying; targets get kicked, obscured, or removed. Having 10 checkshots instead of 7 is better. Having as much control over your survey is better. We're not talking local pile surveys, we're talking multi-million dollar engineering projects. Spending a few thousand dollars is a drop in the proverbial bucket. We're not happy with sufficient, we want the best achievable.

 

[amicron]

I think you missed my point. I never said “millimetric.” Millimeter accuracy is not attainable. It’s less about the level of accuracy and more about a confidence level in claiming and proving a specific accuracy and repeating that accuracy.

We would probably get similar results 9 out of 10 times. But, depending on who is using the data, that one time can get you in big trouble. And you may not be able to pinpoint the problem and fix it.

I’m not saying what you are doing is inadequate. I’m justifying PPK even though you still have to set checkpoints.