Merging projects - misalignment


I am trying to find the best solution to merge projects in the following conditions:

Several flights (typically 4 or 5) are performed in a GPS denied environment and we collect oblique images.

I use Pix4D tagger to get a proper scale and orientation, but I can only put targets in one corner of the project, because the rest of the area is not easy to access.

At the beginning of each flight, the drone captures a few images that contain the targets, but after this there are no more targets to be seen. The flights overlap with each other, so that very similar images can be found in different flights.

Processing each flight independently works well. The targets are well recognized by Pix4D tagger and the project is correctly scaled and oriented. There is probably some deformations in the point cloud but I don’t need a very high precision.

When I try to merge the projects (new project -> project merged from existing project), I observe a misalignment of some projects: for example a single wall produces two layers of points (with 0.5 to 1m shift)

I’ve tried the following to better align the projects:

  • “Rematch and Optimize”

  • Add 10-12 tie points, well spread over the whole area, and making sure that I identify them on 20+ images coming from different flights. Then, “Rematch and Optimize” again.

But it does not improve the misalignment of the projects.

How shall I proceed? The solutions I could imagine are:

  1. Process all images (from all flights) together. Would be ok for smaller projects, but here I have too many images…

  2. Before processing the sub-projects, add to each sub-project a few images from the other sub-projects. In this way, the sub-projects would all share some common images. I don’t know it this would help. If the images have the same name, will they be considered as a single object and become kind of tie point?

  3. Add more manual tie points: theoretically the right way to go, but it wasn’t successful yet.

I have the feeling that when merging projects, Pix4D only considers the position and orientation of the cameras, but does not try to match images of the different projects by their visual contents. I expected that “Rematch and Optimize” would do the job, but it does not improve anything in my case.

How can I get Pix4D to add matches between images of different flights, and recompute the point cloud with these new matches?

Thanks for your help!


Thanks for describing your project. 

Instead of adding Manual Tie Points (MTPs) after merging, I would add them before merging in the subprojects. More specifically, I would add MTPs in areas that are overlapping between two subprojects. The same MTP should be marked in  both overlapping subprojects and have the same name. In each overlapping area I would add about 3 MTPs. 

Then, when you proceed with the merging, the software will ask you if the MTPs and objects in the subprojects are the same and you can click Yes for both. Then, the software will take these into account to align the subprojects between each other. 

I would recommend to read this article on merging as well:

Let us know how it went :slight_smile:

1 Like

I have the same issue.  I have 5 MTPs and ran step 1 before merging.  I get two roof lines on my 3d mesh.  Looks like one project is sitting at a different elevation than the other.



@Kurt, we recommend adding more Manual Tie Points or Ground Control Points in the overlapping area of the subprojects. For more information: How to add / import and mark manual tie points (MTPs) in the rayCloud.

Do you notice some other differences between the subprojects, for example, different values for Internal Camera Parameters?