Post

3 followers Follow
0
Avatar

Doming Effect

Hello,

There can be a technical quirk in contour/DEM generation called "doming", I have attached a reference below.  It results in distorted elevation mapping.

Is this a consideration for MME processing or field data collection?  Apparently, it can be mitigated by using non-parallel flight lines or possibly in the point cloud processing(?).

Any thoughts on this topic as it applies to MME?  I couldn't find any mention of it on this site.  Thanks very much . . . 

Reference:

https://forum.dronedeploy.com/t/considering-doming-effect/892

I would like to know if you are considering the “doming effect” for DEM-generation? It appears, when taking images from near-parallel directions and causes systematic errors in the 3D-model (DEM). See this paper for more information:

Therefore, some convergent images may help avoid this problem. Couldn’t this be done, using the center of the area to circle around and take some additional images?

 

Springfield Harrison

Official comment

Avatar

I am sure there is probably something to that but the way the DJI waypoint system works this would be nearly impossible to do since the flight is straight from point to point. You could likely approximate the curves but it would eat through the 99 waypoint limit to quickly to be useful. Maybe if we had more waypoints...

Zane
Comment actions Permalink

Please sign in to leave a comment.

10 comments

0
Avatar

OK, thanks Zane.  There is also an option to include some oblique images as well for mitigation of the doming.  Could MME handle such a mix of vertical and high angle (20' off vertical) images?

Doing an 80 acre project shortly with the usual lawnmower pattern but might add an oval loop of obliques.

Thanks again . . .

Springfield Harrison 0 votes
Comment actions Permalink
0
Avatar

We have found that "doming" goes away when you use sufficient overlap. Also, it helps to have a feature rich survey area. 

Maps Made Easy can handle oblique images as long as they do not contain the horizon or open sky. 

We do not recommend orbit flight for data collection. The sideways flight causes changes to the shapes of the features due to the camera sensor's rolling shutter. 

Zane 0 votes
Comment actions Permalink
0
Avatar

Right, we'll hope for no doming although I understand that it is related to less than ideal lens correction parameters.

Springfield Harrison 0 votes
Comment actions Permalink
0
Avatar

I'm not commenting on the doming hypothesis using straight lines in particular.  But I believe it is entirely possible to use dji's existing hooks and with the addition of a single waypoint per leg obtain a slight radius.  Litchi has always been able to do this.  They call it "curved turns" and the craft will not fly directly to the mid waypoint.  Couldn't something like this be programmed in MP?

Dave Pitman 0 votes
Comment actions Permalink
0
Avatar

If you read the paper the straight flight is what causes the issue. Adding some turns and going back to straight flight won't likely fix the issue. You would have to do lots of points and then all the sudden you aren't covering much ground. Approximating parabolic flight paths with linear sections isn't going to be super helpful. 

Map Pilot flights do not go directly through the waypoints either since the stop and turn mode is not used. Overlap is far more important to getting a good reconstruction. 60% is not enough overlap which is why we recommend 75% or more. 80% is best. 

It may be helpful to collect data at a lower overlap setting, roughly 50%, and then do another pass over the area after rotating the pass orientation 15-20% to get the effect they are describing. Using a lower overlap setting and then doing the same thing 3 times would be even better. Then 4. Then 5. Then 6. Then... More pictures is almost always better.

It is hard to back out what the resultant overlap of what they are collecting turns out to be. It is likely higher than an equivalent of 60% so comparing the outputs using this technique with grid flight at 60% overlap is an apples and oranges kind of situation. 

Zane 0 votes
Comment actions Permalink
0
Avatar

The offset passes makes sense, but there is always the chance (by chance) that a particular spot might end up with insufficient overall coverage, yes?

The paper also suggests adding slightly oblique imagery and you guys have said that that doesn't work too well with MME's processor.  It's tough to know what is the best approach sometimes.

Dave Pitman 0 votes
Comment actions Permalink
0
Avatar

Dave:

You can upload oblique images to MME as long as the images don't contain the horizon. This is covered in the Data Collection guidelines. 

Offset like this:

Zane 0 votes
Comment actions Permalink
0
Avatar

Thanks Everyone,

My understanding is that a combination of straight flight lines and inaccurate lens parameter modelling causes the dome effect with elevation.  Perhaps some processors have more accurate lens data than others (?).  Otherwise, the fix seems to be avoiding straight flight paths, one way or another.

Be cause the flight planning software already does amazing things, adding in parabolas or s-curves while still maintaining overlap should not be impossible, although probably challenging!

Springfield Harrison 0 votes
Comment actions Permalink
0
Avatar

Zane, Yes, that is what I thought you meant.  I was just saying that with that method, and using less than optimal overlap on each mission, that even though when the image set is combined and the total overlap should be good, there is the possibility (a random chance) that some spot within the subject area may end up with not enough overlap because the subsequent mission doesn't know where the images from the first mission were taken.  Its probably likely that over all overlap will be sufficient but there would be some "chance" involved with this strategy, correct?

Dave Pitman 0 votes
Comment actions Permalink