Support Website Contact Support Blog

Misalignment at edges between merged subprojects

Dear Community,
I have a bigger project of ca 22 000 images (1.9 TB) which I am trying to process into one orthomosaic and digital surface model (DSM). I split up the project into 12 subsets using Pix4D with ca 2000 images in each subset. Then my workflow was:

  1. Do initial processing with all subsets
  2. Merge projects:
  • Add Manual Tie Points (MTP) in Project A
  • Add the same MTP in Project B (name them the same)
  • Reoptimize in each project
  • Create new project with the option “Project Merged from Existing Projects”
  • Choose to treat objects with the same name as he same
  • Click finish
  • Reoptimize
  • Further Processes

I tried to merge 5 of the subsets and run all steps and when I started to inspect the dense point cloud I realized there are some problems. Sometimes at the border between the subsets there is a misalignment and a gap in the pointcloud:

Here we can see an area between to subsets:

If we zoom in we can see that the images were taken under different light conditions:

And here the misalignment becomes visible:

Here is another example at the road:

I guess this misalignment will cause artefacts in the DSM and since my goal is to to DSM analysis I would like the DSM to have as high accuracy as possible.

Any ideas what can be the problem here and how can I solve it?
Are the images taken in sun too different from the ones taken under cloudy conditions?
Would it help to add more MTP in the errorous areas?
Should I better perform all steps on each subset and then merge them in another program?

Btw I cannot provide a quality report because Pix4D crashes every time I try.

Thanks in advance :slight_smile:


22000 images! wow! Very challenging! Maybe you want to give it a try with our new product Pix4Dmatic that is optimized for very large projects. It is possible that Pix4Dmatic processes it in one project.

Regarding Pix4Dmapper, adding more common manual tie points (MTPs) especially in the areas with different light conditions will definitely help.
Here you can see how MTPs helped with a different light conditions dataset:

Also, merging 2 projects at a time is recommended (instead of merging 3 or more projects at a time)

Guessing your errors come from reflections due to sunlight. I find that sundlight (especially a strong, but low sun) make a lot more noise in the pointcloud.
However I, probably like you, don’t have the luxury to choose when to fly. Generally I find that the actual gound is at the same hight, so you just have to remove unwanted noise. This is ofcourse rather a pain for you with that amount of data. I wish there was a tool for selecting an area, thell the program that this is a mostly flat or uniform hight/incline area, and have it remove the worst noise.

I don’t think Pix4Dmatic will slove the noise issue, as it does that same as Pix4Dmapper. The issue is in the limitations of the software as it needs similar lighting conditions.

That would be a very useful tool!

Hello Christina,
I think I heard somewhere that Pix4Dmatic is not yet available for 16-bit-images?

I tried to merge only two subsets with additional MTP at the road part. Here is a screenshot of the point cloud in Pix4D:

in the attachment you find the quality report of two merged subsets with MTP added to the road part. In the preview there seem to be a lot of noise as described in the link you sent. And there is also a clear line artefact along the edge between the subsets:

So adding MTP to noisy or errorous areas doesnt seem to help so much and would be very time consuming if I have to do it for all subsets.
Any processing options I could change maybe?

Merge_5_10_report.pdf (2.2 MB)

I think I heard somewhere that Pix4Dmatic is not yet available for 16-bit-images?

Here you can see the compatible cameras for Pix4Dmatic:

Thanks for sharing more info! It helps us to better understand the case.
I suggest you process with the 3D Maps template:

That will process with Full image scale and some other options that I believe are better for your project such as Use MTPs


How accurate are your images’ coordinates? Do you use an RTK drone?

Also, I suggest you mark some of your MTPs on more images. Some are marked on only 2 images. I recommend min 5 images

Why don’t you change how you split the projects? Splitting by flight results in different radiometry on adjacent projects which is giving you the problems when merging. Try the opposite approach. Use half of the images from two adjacent flights in a single project. Try to split the projects within the same flight so the radiometry at the adjoining edges is the same.

Pix4DMapper may be able to handle the large radiometric variability better within a single project as opposed to across projects.