I would like to know if what I am seeing is an error of the image processing. In the elevation maps the edges always remain at a lower altitude than the center of the map. It is as if the map is always curved. I give some examples so that you can see them. In the 3D model you can also see this effect.
Hey!!! No feedback??? Anybody??
Since you are looking at elevations, are you using GCPs? I know that as you reach the edge the data has more potential for error due to less image coverage. I typically would add another pass outside of the subject area to ensure that I have the overlap. In addition, what were you flight parameters?
So i took a look at your elevation and models and nothing looks terribly off, however not knowing what lies outside of the mapped area doesnt really let me see if this slight curvature is due to the actual ground curving. I would agree with @dragonflyAS it is always best to capture data that is 10-20% outside of the area where you are looking to produce quality data. This will help to insure enough data points for proper tie points to be created. If you are lookign for a tighter absolute accuracy than having at least a few GCP tagged with an RTK GPS will let you verify your map a lot better.
I would also take a look at your resolution (how many cm/pixel is the map done at, and the accuracy (RMSE, X,Y,Z) remember that if you are not utilizing a drone with RTK or GCPs with the map your vertical accuracy is not going to be very good ( passable but not survey grade) As altitude is notoriously hard to accurately extrapolate from GPS.
Also as i am thinking about it, remember that the elevation is being determined from the relative point of take off for the flight, not in an absolute datum like (NAD83).
Hi, thanks for the feedback!!!
I’m not using GCP, I hope DD can give me a correct elevation map without spending any more money besides the license. I know the precision will not be exclente but i hope the map has to be, al least, a decent map.
I do not think the problem has to be with overlap. In the first map the altitude was 130 meters with 65% of sidelap and the second map was 65 meters with 75% of sidelap. But the edges of the map is where the drone take more photos because it brakes and takes the curve while still taking photos so the overlap is higher at those points.
In addition, if you see the maps, the elevation does not suddenly fall on the edges, but is descending gradually …
RMSE 7,2 m (4,9m ; 3,7m ; 3,9m)
RMSE 5,5 m (4m ; 2,1m ; 3,2m)
You are correct that even without the GCPs you should be getting a fairly accurate elevation (still remember the elevation will have the most error in it height wise) and using GCP will in crease your absolute accuracy but really shouldn’t have a huge affect on the resolution. I Tend to bump my side and front lap up some more, but again should not really cause a huge issue. One thing you could do to increase the elevation accuracy is to fly at two heights for the one map, giving it some more data to work with. But lets go on the assumption that every thing on this end is fine and look at the maps.
A couple of things, since this is relative it would be good to know what the range of elevations in the map are (meaning what height is the darkest blue and what height is the red) While the drone does in theory have some more overlap in turns, that doesn’t actually mean you are capturing any new data. [i am unsure if they take based on time, or actually based on positional overlap, i would think the second but unclear]. Really what i mean to say is to map a larger area so that you ensure anywhere you are trying to take data from is well within the bounds and wont suffer from decreasing accuracy at the edges where tie points invariably are less. In fact give that a try, in crease the mapped area and re-export to see if that decrease of elevation is really there. Also check the difference, it might be a difference of only a few cm, or a meter.
Thanks for the tips, i will use them in my next flight.
The max elevation of the terrain in map 2 is 6,87 mts, and the min is -3,04 mts.
I used a P3 adv with a 4K cam, automatic sets.
Hmm, so that’s a 9m difference between high and low points (could be a little exaggerated due to tress, ditches, etc. ) P3 has a 12 MP camera on the 4k advanced so that should be alright. Interesting, yeah give those a try and see how the next ones come out. hope this helps
I am having the exact same problem. I am testing Dronedeploy and all my 3D and DEM images have a curve. I was researching about the problem and it seems that is due to the camera lens. On some image processing programs such as pix4d desktop edition, you can set your camera in order for the images to be corrected. Do you know if DD has the option to set the camera before processing? Or any ideas on how to correct this?
Btw, I set my account settings with “Use 3rd Party Camera = ON”
Thanks in advance!
Gonzalo/Remotely Possible: What you are experiencing is the well-know systematic SfM doming (elevation) error experienced when the imagery is captured using traditional (linear/parallel) flight lines. This is caused by accumulation of the lens distortion error through the image block assembled during the SfM workflow.
To avoid this error, it is helpful to fly curved flight lines which mitigate the doming error. Google “Gently Curved, Convergent, Non-traditional Drone Flight Lines”. Also, take a look at the following reference material.
Minimising systematic error surfaces in digital elevation models using oblique convergent imagery
First published: 16 March 2011
Results of the simulation process, the laboratory test and the practical test are reported in this paper and demonstrate that an oblique convergent image configuration eradicates the systematic error surfaces which result from inaccurate lens distortion parameters. This approach is significant because by removing the need for an accurate lens model it effectively improves the accuracies of digital surface representations derived using consumer‐grade digital cameras. Carefully selected image configurations could therefore provide new opportunities for improving the quality of photogrammetrically acquired data.
I hope this helps.