Elevation Data using Mavic Pro


I have flown two missions in this past week and both have elevation errors. It is just creating either a dome or depression out of the entire map. The first I thought was just an issue because I had a battery swap and maybe something got funky. The second was all flown on one battery. You can view it here. Both were flown with iPad Air 2 using a Mavic Pro.

First Flight

Second Flight

I’m not sure, so don’t quote me, but doesn’t the Mavic have some form of fisheye lens? I notice under the specs on DJI’s site, that it lists it has some level of distortion on the lens.

Lens	  FOV 78.8° 28 mm (35 mm format equivalent) f/2.2
Distortion < 1.5% Focus from 0.5 m to ∞

I imagine that could play a role in your problem, however it may not be the only issue.

Well, I might research that a little more as I didn’t know that to be a possibility, but I have successfully created a good elevation map with the Mavic. I do not notice fish eye when taking stills. I created this one using my Android Phone.

Good Elevation

Yeah, I’ve seen images from the mavic as well and didn’t really notice distortion so, it probably is something else.

This is not in any way an issue with your Mavics camera. This is an intermittent problem that we have seen with a few maps that generally share the characteristics of not having a large elevation change throughout the map. We are working to resolve the issue but because it is so intermittent it has proven more difficult than most.

To answer your question about the lens on the Mavic. It does not have a fisheye lens. It has a similar Feild of View to the X5 camera. With a slightly wider FOV of 78.8 compared to 72 on the X5. A comparison of all the X series cameras can be found here. One thing to note is that if you do have a longer lens your FOV will decrease. This means it’s a great Idea to fly higher so you can capture more data that can be used during the stitching process. All senses will have an inherent amount of distortion associated with the camera.


I just flew my first mission with a DJI Mavic Pro and experienced the intermittent problem you mentioned. The actual elevation change in my mission was around 15 feet (215 to 200 feet above sea level) yet the Drone Deploy elevation chart and 3D map show a dome with a elevation of from 41 feet to 206 feet. The drone flew at a constant 246 ft AGL for the mission. Do you suggest flying the mission again? Flying at a higher altitude? Oddly your software apparently interpreted my files as coming from an Inspire - https://www.dronedeploy.com/app2/data/1491586226_SAMINSPIRE

Any update on this issue DD?

I have had this happen a number of times now mapping quarter sections for farmers.

Completely spitballing here but would this have anything do with curvature of the earth?

Thanks for the response. Sadly, it appears to be an intermittent software issue (the worst kind in my book) on the DroneDeploy side of things according to Zack1 who sounds like he may work for DD. Elevation Data using Mavic Pro Hopefully they are going to resolve the issue in the near future.

Thought I’d throw my two cents in here:

This same effect can be seen across most Structure from Motion software platforms. I have personal experience encountering doming (as it is called) in Maps Made Easy, Precision Hawk, ENVI OneButton and Drone2Map.
For more information, check out this article: http://onlinelibrary.wiley.com/doi/10.1002/esp.3609/abstract

Drone Deploy, I think we’d all love to hear what progress you have made in mitigating this issue at this point in time.

Thank you!

Systemic Bowling Effect on UAV produced maps is something that is very difficult to nail down and everyone in the industry has run into it at some point as @jread5k pointed out.

I have read that if a few oblique images are taken during the flight this can help but there is no silver bullet at this point. I do know that our engineering team is working on the issue and have made some tweaks to the map engine that should help. I personally have not seen nearly as many “doming” or “bowling” within agricultural maps which were the main groups affected last season.


I am new to DroneDeploy and was disappointed that my first two (agricultural) maps show pronounced doming, swamping any actual ground level changes. My third mapping may have the same fault too but since the area being mapped had a real change of 100 metres maybe the real world swamped the processing error. I have not tried adding an orbit to one of the previous maps though I see a suggestion above that might correct it. In fact I don’t see a way to add an orbit after the event. I would have to copy to a new map and lose one of my remaining free trial maps to find out if this is a fix or not!

Hi @aj_uk,

Please check your inbox on the forum. I’ve sent you a private message regarding your issue.


I’ve just had an issue related to this as well. Drone Deploy seems to have an error in ground elevation on a recent map, that reversed the slope of a slab on a construction site I surveyed. I ran it through Pix4D to double check and I’m glad I did since it could have stopped work on his site. This is a big problem for those looking to use this app professionally.

Take a look at the following publication - https://authorcentral.amazon.com/gp/books/book-detail-page?ie=UTF8&bookASIN=B07CWJR2L5&index=default - which is written specifically to provide gently curving, convergent, non-traditional (not straight/non-parallel) flight lines that do not allow the lens calibration errors to accumulate via the processing of images (image block) captured using traditional long, straight, parallel flight lines. I am working to provide a flight planning app (web-based) to support curved flight lines.
Jim_Dow (jimdow528@gmail.com)

There shouldn’t be any need to create a non-standard flight pattern for a condition that is a very rare occurrence and is all but isolated to older drones. There are two minor issues with a curved flight pattern for mapping. One being that your drone and GPS module and compass are in a constant state of rotation, no matter how minor, and this is not good for accuracies. You wonder why slow and steady and straight lines are more accurate… Another issue that you will run into issue trying to create a path that is truly parallel that creates a consistent optimized overlap. Overlap is what the whole scenario is about so if you end up with less overlap because of the angular relationship of the images you will get less tie-points and thus a less accurate model. I tried this in Litchi before because I had a site with a shape that the lawnmower pattern just wasn’t very practical on and the stitching and horizontal accuracy was not good at all. I ended up having to refly the mission and just deal with the extended flight.

Michael: Thanks for your response.

Jim Dow

Michael: My experience indicates that the SfM dome effect error is very real; and it occurs frequently, with imagery collected with both piloted aircraft and drones (older and newer versions).

Also, for most all recorded flight tracks, the GPS/INS instruments are in constant translation/rotation motion regardless of whether the flight path is straight or gently curving and convergent. Recent research indicates that gently curving convergent flight paths do mitigate the accumulation of lens calibration errors in the SfM workflow.

Image overlap is not an issue with gently curving convergent flight paths.

Thanks for your response.

Jim Dow

(205) 461-5747

I never said that doming wasn’t real. As a matter of fact I have stated before that I experienced it myself with the Phantom 3 Pro. If you look at the records on this site it rarely occurs with the newer drones and is normally due to the fact that people are flying too low with not enough overlap. The other factor that can cause doming that is hard to get rid of is when it is a very large site. I find that anything over 50 acres needs more attention.
You are correct that the drone is in a state of constant recalculation, but when you put in a curve it is an added computation that virtually every part of the Drone system has to recalculate for. More variables equals more calculation equals inconsistency. Case in point, there are several companies out there that I’ve done research that states that running a survey mission in headless mode is more consistent than turning. Most of them mention it specifically in regards to camera error and GPS fix.

Look up an Amazon eBook entitled “Gently Curved, Convergent, Non–traditional Drone Flight Paths” for a solution.

Make sure to fly high enough. Particularly with the Mavic series as they have a lesser field-of-view. This is even more important when flying over homogeneous subjects.