Author Topic: Detect coded target and use only them as tie point for the alignement step  (Read 1948 times)


  • Newbie
  • *
  • Posts: 1
    • View Profile

I'm french engineering student [sorry for my english]

I work with metashape on the purpose to extract the coordonates of coded targets in a local system / machine system, when the coded targets are abundant on several photos.

Actually, I use metashape with the alignment function, and after alignment is finished, I detect coded targets and I can extract my coordonates in a local system.

Alignment photo is doing by "bundle adjustment", using detecting and matching a lot of tie points between the photos. This step takes times, I want to reduce it.

I want to detect coded target and use only them as tie point for the alignement step. Do you think is it possible with Metshape ?

I saw that it was possible to code python scripts. My idea is to create a list containing the coded targets to consider them as tie points. Then, to ask the software to align the photos using only this list?

What do you thing about it ?




  • Hero Member
  • *****
  • Posts: 1347
    • View Profile

I think that you cannot align photos just using detected coded markers as tie points. Maybe Alexey can confirm this. This would be like just doing alignCameras() without previous matchPhotos() using only detected coded marker projections....

A workaround would be to do an alignment with low/medium accuracy and low Keypoint limit (1000) which would be very fast, Then detect coded markers and do optimizeCameras(). To put a greater weight on your coded marker measurements relative to automatic Tie points measurements, you could set Marker accuracy to 0.2 pix and Tie point accuracy to 1 pix in Image Coordinates accuracy section of  Reference settings window as shown...

J'espère que cela te sera utile!
« Last Edit: March 19, 2022, 01:30:15 PM by Paulo »
Best Regards,
Paul Pelletier,


  • Newbie
  • *
  • Posts: 32
    • View Profile
    • 3create

could you post an example image with the coded targets on the object you are reconstructing?