My main need is to measure the coordinates of the GCPs that I place on the ground before flying with the drone. On rare (very rare) occasions I also need to take direct measures to materialize in the field points of which I know the coordinates “on paper”. Having absolute coordinates would be the best but not a priority, even relative ones would suffice because I am not interested in global georeferences but I just measure in order to obtain surfaces and volumes.
I don’t need “extreme” precisions. A few centimeters of error are fine. The Z axis is important, if I get relative heights wrong of meters its a disaster.
The areas to be surveyed are usually in open countryside or mountains, with no vegetation at all; but the orography is very irregular with very violent natural changes in altitude (often mountain fronts).
Around me I have 3 fixed stations, belonging to the EUREF Network, at the following distances: 240 Km, 280 Km, 340 Km. I NEVER have any known points in the field. The terrains to be surveyed are small (never exceeding 80,000-100,000 square meters).
This is the scenario. Now the questions.
In these conditions, do I need a base and a rover or can I make do with just one receiver ?
Is an RTK or PPK approach preferable in my case ? Which of the two is more immediate and faster ?
If I buy a receiver with only the L1 band how long does it take, on average, to get the coordinates of a point ? How long does it take to do the same with a multiband receiver L1, L2 and L5 ?
Your EUREF base stations are really far for RTK, I think you need both base station and rover to do precise measuring. You can establish your point, if you are doing multiple surveys, with your Reach RS+ base station and mark it somehow. Keep in mind, that for establishing a point you need to keep base station for quite long period of time (up to 6-24 hours). Mountain fronts are really tricky, cause you don’t have almost a half of the sky view.
I would go with PPK, because it can be a problem to maintain strong reliable connection with the base station in these conditions. RTK is more immediate and faster but require a good connection between devices. If you will have straight vision with the base station all the time, you can try RTK after establishing base station.
From my experiments, I would say it depends on the conditions and on a place, where you put the base station. I would start with 2 hours of data collection and after it I would try to stakeout the base station point. L1, L2 and L5 will do it faster, I suppose, but I don’t have any numbers on it.
I can, almost always, place the base in a point with full sky visible and with a full view of the rover.
The rover, at some points, may not always have the whole sky visible.
So, I got two workflow ideas.
I arrive on the site and I place the base in any point with a full sky visibility. Then I turn on the base and I start recording raw data. Immediately after I also turn on the rover and start recording. After about 15 minutes I place the rover on GCPs and hold it for 2-3 minutes at each point.
When I get home I download the data both from base and rover. I post-process the base data with EUREF base in order to get a “somehow” position. I post-process the rover with the base on field (including the coordinates obtained from the first process).
I arrive on the site and I place the base in any point with a full sky visibility. Through a smartphone GPS app averaging I measure the coordinates I would assign to the base. Then I turn on the rover and start measuring the points getting the base corrections.
I think, both approaches will work. Anyway, by using both ways, you can get raw logs out and post-process them. If you really need very precise vertical differences, I would go rather with theodolite, laser tracker or total station. Vertical differences with any GPS are quit tricky.
I quickly read over your workflow. If you understand “relative” measures and you know that’s what you want, absolutely go RTK. You don’t need multi-band for the scenario you’re describing, in my opinion.
You also don’t need to use a mobile app for averaging position. Reachview base will calculate it’s position by averaging for up to half an hour. This will be perfect for your “somehow” position. In this scenario, you can have your base logging the whole time as well as your rover. If you have trouble getting a fix somewhere, you can still post process that point and just let your rover stand there for fifteen minutes. Getting the initial fix can take several minutes, but after that you will be surveying RTK “live.” It should maintain the fixed solution everywhere you go, as long as you keep your rover facing the sky.
As far as the vertical error, you should have nothing close to 10 cm of error.
Because you had your base logging the whole time, you can post process later with official data and shift your whole survey if you someday required absolute accuracy for a survey.
When I talk about relative measures I mean that I am interested only in the dimensions of the what I survey. Just distances and heights to be able to measure surfaces and volumes. I don’t care that the topographic relief is well positioned in the world.
So, if I understand correctly, I just need to position and turn the base on and let it calculte its position. Then I have to wait for an initial RTK fix (like 15 minutes) and then I can go around with the rover and get the position of each point I need (without waiting on each point)
Furthermore, if one day I wanted to georeference the topographic survey, I could use the raw data collected and post process it with RTKLIB (PPK mode).
All correct ?
What could be, on average, the error in measuring relative heights? I don’t care about the absolute height but that the physical heights of the object I detect are correct between them.
Is there a precision difference between the Reach RS + and the Reach RS 2 ?
Yes, all correct!
I can’t give you precise measurements but somewhere I read that normal vertical error is twice that of horizontal error. I would expect the max vertical error to be 5 cm but probably usually within two or three centimeters, maybe less. Someone want to help us out here?
As I understand it there is not a significant difference in accuracy between RS+ and RS2 in a normal working environment. RS2 will get a fix faster, will hold onto it better, and the distance between the base station and the rover can be significantly greater. But a true fix on either of them will still be centimeter level accuracy.
Bear in mind, that relative altitude differences while using UAV are mostly produced by photogrammetry and not GPS data. GPS can correct altitudes and GPS data is mostly useful on the sides and far corners of your UAV data. Once I made an experiment with UAV processing of photogrammetry data. I made a survey of relatively flat territory with small articulations in terrain (max. 2-3 m of difference between different terrain areas). I made the same processing on 8 different PCs with identical parameters. What I got, were 8 different DEM surfaces. When you compare afterwards generated DEM values and measured GPS points I got different values from all DEMs. Differences sometimes were up to 16 cm, but in average around 10 cm. I assume, that differences between DEM points not covered with GPS measurements will be also around the same error. In Agisoft Photoscan every processing will generate unique model, because of random factor involved in picture alignment. That’s why you get these differences.
Error in GPS points is also calculated based on measurement count. If you do fast measurements, your sample will have relatively smaller amount of measurements, than longer acquisition times. So, you will get also smaller error, which doesn’t mean, that you measured precisely. It means, that you don’t have too much measurements to calculate error. Your results will not be significant.
Also, bear in mind PDOP. Sometimes, you can get lower error, but your PDOP value will be high. In mountains it is very easy to get high PDOP, because of obstructed sky. Even if you will have your local base station on the mountain with full sky view it doesn’t mean, that your final measurements with rover will be precise, if you are measuring in the valley. Most probably you will see satellites, that are positioned between 2 mountains. These satellites will be arranged in something close to a row and that’s when you get higher PDOP.
At the beginning I believed that the GCPs only served to georeference the model and that I could do without them for the relative measures. In the first (experimental) flights I did the drone stopped in the air whenever it had to take a picture. The models that I obtained with this system, without GCPs, had very low errors in terms of distances and heights. Unfortunately flying in this way consumes a lot of battery
So it is true that the errors in the relative measurements are due to the camera’s optics rather than the GPS. The more the photos are taken in “static” conditions and the smaller the error becomes, the faster the drone flies and the more this error increases. The GSD, on the other hand, does not affect this.
Another idea that I had was to replace the GCPs with a sort of “scale constraints” … often the stuff that I survey is characterized by sharp edges and vertical walls (stone quarries) so my idea was to measure some axes, both horizontal and vertical and impose these measurements inside the Photoscan process… I’ve never tried this…
In RTK mode, once I get the fix, I should be able to survey points quickly. Am I wrong ?
Yes, you will survey quickly, if your FIX is stable all the time. For GCP I usually survey a point for 20 seconds in RTK. For things, that I don’t need very precise, like GPR profiles and electrical resistivity sounding sensors I do for 5 seconds.