Doming and the DD's dropping of 3D mapping for DJI P3's

Anyone who tries to make large maps using Drone Deploy with a DJI phantom will find beyond around a hectare, that the elevation image will have an unacceptable ‘doming effect’ on the image, that the centre of the map can be up to 2-3 metres higher than the periphery. This is a well known issue, based around issues of lens asymmetry, and cannot be easily corrected. The solution is to take oblique photographs as well and DD had a useful feature until late last year, where the drone would circle and take obliques at the end of the survey.

Alas DD have now dropped this is their latest release, with something called 3-D beta. I am sure it may be better, but we cannot try it, as this feature is inexplicably turned off for all the DJI Phantom 3 drones. This means that DD is now practically useless for all serious users of P3 drones, who want to generate DTM’s and .las images.

It is very simple change please can you reinstate the circular oblique function - it worked well (PS I was filming with Discovery Week, and it was deeply embarrassing that I was unable to produce the promised map)

Mark Horton

I still fly my P3 Pro and haven’t seen any of this “doming” you speak of, but I’ll keep an eye out for it. As for the circular oblique function, I tried that once and the camera was pointed level with the horizon and, from what I recall, I wasn’t able to change that. I do like the idea of the feature, and the one for taking obliques around the outside edges of a rectangular shape as well.

Hi Mark & Marc!

I’m curious to know if by “doming”, you are referring to the “bowling effect”? The bowling effect is a photogrammetric issue that typically affects the elevation map. Unfortunately, the only known workaround for it at this time is to capture oblique images and add them to your existing map. We’ve noticed that adding obliques can help reduce the likelihood of the bowling effect.

We replaced “Orbit at End of Mission” with our 3D Mode, which is designed specifically to generate crisper, more detailed 3D models. We understand that it’s only available for newer DJI drones, but you should still be able to capture oblique images manually or with DJI’s POI mode. You can always add these additional oblique images to your dataset to offset the bowling effect and/or to improve your 3D model. These are two great alternatives for those flying with the P3 series!


The only time I have experienced “doming” is when the units were incorrect thus the scale of the elevation change was exaggerated. That being said, we use GCP’s on every mission so elevations are spot on.

Thank you Stephanie for the response. Yes doming = bowling. Oblique photography is a workaround, and your obit at the end of mission neatly mostly resolved it. So we were very disappointed when it got dropped in your latest release - and the alternative isn’t available for P3 drones - where there are thousands of us out there using P3’s - and will be doing so into the future as its the only affordable and compact drone where we can change cameras for multi spectral use.

The trouble with your suggestions are 1) to do it manually over a large area is practically impossible and very time consuming (as the effect is on large area survey this IS a problem) 2) POI works for small areas, not the 1km + square areas that we are doing 3) Critically it requires landing the drone, swapping to DJI Go and relaunching. This means extra flight time, more batteries and possible changes in light conditions (here in the UK cloud cover moves fast). Some of our large surveys use three batteries, so this is a major limiting factor.

So a plea from your loyal users - please reinstate orbit at the end of mission!



Hi Mark,

Thank you for taking the time to explain your use cases and why manual obliques or POI might not suit your mapping needs. I understand your frustrations but unfortunately, we currently do not have plans to restore “Orbit at End of Mission”. We discontinued the orbiting feature because it was not performing reliably for all drone models and it had known issues with the DroneDeploy app. I’m sorry for the inconvenience. I will be sure to communicate your feedback to our Product team for future consideration.


If you know what angles to take your pictures from shooting the proper isometric images is not very time consuming at all. Try strafing in addition to orbiting. Do multiple POI missions evenly spaced across the projects. While others may not recommend it I never need to land the bird when swapping from DD to DJI Go or Litchi. if your settings are correct it won’t do any but hover.

Can you provide a link to such a project or even share some of the images? I am interested to process them a couple of different ways to isolate the issue.

I experienced this several times when I was flying the Phantom 3 Pro. An extra flight with low overlaps at an altitude of 300 ft or more stopped it. Crosshatch flights also work, but are much more time consuming. The second flight method seemed to be much more efficient.

The “doming/bowling” elevation error can be associated with an accumulation of lens calibration errors during the SfM workflow - especially when traditional (linear/parallel) flight lines are used. Adding oblique images at various altitudes is recommended as a workaround, especially when the oblique photos are taken using gently curved, convergent, non-traditional (non-linear/non-parallel) flight lines.
Google “Gently Curved, Convergent, Non-traditional Drone Flight Paths”. The technique works for small and large areas as well as for long narrow corridors.

This works unless you’re using gcps. Gcp’s plus obliques equals bad idea.

Michael: Matching GCPs is the whole idea! Unless you can ground truth the horizontal/vertical dimensions, what do you have that is useful?

Simple stakeout inspections with survey grade GNSS systems and it’s pretty straightforward. Neither the RMSEs or real world values were as good with obliques. Structures may look better, but I have little use for them in our actual process. They get wiped off anyhow.

In our precision agriculture application, the doming elevation error is 7’ over a 2500’x1300’ field. This turns out to be of little consequence because we construct a DTM in Rhino from the ground visible between all the rows and columns of produce. After subtracting the surface made from the DTM from the 3D model, the residual ground variation is about 0.1’ (1.2"). This enables accurate measurement of the produce height above the local ground level. Flying curved flight lines and taking oblique photos would increase the mission time, require 2 batteries and not improve the results unless there is something I am missing.

But the doming effect is quite real and consistently shows up in DroneDeploy, Photoscan and Pix4D for these 80-acres sites.


Can someone please try a practical flight with the curved flight lines? I would, but I don’t have a P3 anymore. Easy enough to plan in Litchi. GCPs would be good in this case.

McCall%20Glacier%20Log-Sprial%20Trajectories Fairbanks%20FODAR%20Traditional%20Flight%20Plan

Take a look at the two attached images to make a fundamental comparison of the curved flight lines as compared to the traditional linear/parallel flight lines - and this example does not even illustrate the most effective approach available when curved paths are connected into efficient loops.

A seven foot doming elevation error (if I were your paying customer) is unacceptable - especially when there is such an easy way available to solve your problem. Refer to the research noted below.

Minimising systematic error surfaces in digital elevation models using oblique convergent imagery

Rene Wackrow

Jim H. Chandler

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.

1 Like


The attached image illustrates curved flight lines for an open pit copper mine in Alaska covering approximately one square mile. These curved flight lines can be flown in complete loops; or they can be flown progressively end-to-end. Now that’s efficiency - no wasted looping turns at the ends of any traditional (linear/parallel) flight lines.

1 Like

For my application I would still have to make a DTM to eliminate real elevation differences. So there is nothing to be gained in my case by flying curved flightlines.

SolarBarn: You now think you have nothing to gain??? You could still capture nadir images (along with convergent oblique images); and your surface definition (as well as your DTM) would be correct - or at least the doming (elevation) error would be minimized!!! The curved flight lines also allow you to minimize the number of GCPs utilized. Now, there’s some real direct cost savings in terms of dollars and time.

We do not need GCP’s as the crops are on a fixed pitch which the photogrammetry captures correctly. If the surface definition is not perfect, the DTM eliminates its impact. You need to keep in mind that I do not need a DTM with correct elevations. Rather I need a surface generated from the DTM that conforms to the earth next to the crops and provides smooth, continuous continuity between these patches of exposed earth. LiDAR people call this a bare-earth model. This bare-earth model is then subtracted from the domed 3D model to create a model where all elevations are relative to the local ground at 0 elevation. This model looks perfectly flat at 0 elevation except for the crops that stick up and whose height can be measured directly as if you were standing next to each plant with a measuring tape. The actual ground may wander up and down a few feet over the 1000’s of feet extent of the site, but every measurement of the crop height is relative to the ground right next to the plant. All of the large scale fluctuations of the ground elevation, whether it comes from doming or actual ground elevation changes, are completely flattened out in this derived model. Tomorrow I will attach some elevation maps showing the before and after site models and the bare-earth model created from the DTM.

SolarBarn: Thanks for your explanation. I understand that you are indeed extracting relative measurements between your “bare earth model” and your crop surface model; and the doming (elevation) error in each model might be expected to be consistent between the two models.

My only point for you is that “Why would you want to have to explain the consistency of the doming error in both models when you can minimize the doming error from the beginning by simply using gently curved, convergent, non-traditional flight lines while capturing some oblique photos along with your nadir photos?” Also, I definitely understand why you don’t use GCPs!