Emlid

Community Forum

Accuraccy of Reach on UAV - GSD


#1

Hi all,

Has someone obtained RMSE errors (X-Y-Z) of 1GSD or less using Reach mounted on UAV comparing against GPS surveyed points on ground?


(Christian Grüner) #2

I think it almost impossible to get 1 GSD accuracy on a UAV with current technology.

You have to be sure to have a mid-exposure trigger (something rare on most consumer cams/drones) and you have to take into account offset from attitude (and the delay in getting that data) and so on.

Of course, things like minimizing distance between antenna and non-parallax point of the camera makes it easier, but comes with its own set of challenges.


(Michael Lambert) #3

1 what GSD? Inch, cm? I routinely get 0.7-1.0in GSD. RTK/PPK has nothing to do with that. Are you talking about the average RMSE of the images as a whole or GCPs? These are two different methods. If you are talking about the map with no GCPs then you should be getting 2-3cm or better.


#4

My question was because we did a flight at 9cm GSD with a Reach on the UAV and obtained image coordinates with Rtklib.

We surveyed 40 points on ground and using all of them as Check Points their RMSE error is 5cm X, 7cm Y and 10cm Z. Using for example 6 points as GCP and the others as Check Points the error is less.

That are very good results but also strange for us because we thought that the error couldn’t be less than the GSD (9cm).


(Michael Lambert) #6

What were your flight parameters? These numbers seem really high to me. I typically don’t fly any higher than about 90 meters, but even at higher altitudes the GSD is usually under 5cm. Did you post process your data? With what software?


#7

@chascoadmin we flight at 400m, with a Sony RX100 II with Hot Shoe to get 9cm GSD, 70% side overlap. We processed the Reach PPK coordinates with RTKpost (all camera positions fixed) and the data with Photoscan.

All the examples that we found with Reach on the forums were at lower GSD, between 2-4cm, so couldn’t compare with them.


(Michael Lambert) #8

Where are you that you are allowed to fly that high? Or are you a government contractor or have a waiver for that? I personally don’t care, but I would not publish it unless you carry aroud an extra $5k…


#9

Yes, being registered uav pilot in my country, you can get special permissions to fly above 122m specifying location and time of the flight.


(Michael Lambert) #10

Great, just wanted to make sure. Thank you for setting a good example. Front overlap? Doesn’t really matter, just curious. Their GSD’s are lower because of the altitude your are flying. When I fly to our limit of 122m I have trouble getting better than 4-5cm. GSD is not an accuracy value, but a factor of the resolution of the camera and your altitude. I would do whatever I could to lower the altitude and accept the use of more batteries.


#11

Front overlap was 80%. For the accuracy and resolution needed for this flight, 9cm GSD was enough, and we surveyed 40 points on the ground.

Independently, we use Reach on the UAV to test it’s performance and we are very happy with the results.

We use another Reach as base and had a Permanent Station near too.


(Michael Lambert) #12

It sounds like you are getting the best you are going to get at the elevations that you are flying at. Anyway you could share a link to the map? Sounds like a very interesting case.


(Ryan Mc Gowan) #13

This is definitely camera related. Photogrammetry needs good images to get accurate measurements. You pixel size (GSD) is going to be roughly equal to your horizontal accuracy. Worse if there’s motion blur or your camera has a rolling shutter. Fly lower in areas that you need more accuracy. Higher in areas where it’s not as critical. Keep records of your altitude and RMS values for future planning.