Show Posts
1
Bug Reports / Clouds from the same dataset don't align
« on: October 23, 2019, 05:40:16 PM »
Hi!
I'm doing some testing with the latest version of Metashape Pro, and found a problem that I can't understand.
I have a set of 83 images and 6 GCPs, which were manually identified in all of them (more than 20 images per point). I generated the sparse cloud with the Lowest alignment (generic and reference preselection enabled) and the dense cloud in all categories in sequence. Then I generated a new sparse cloud, with the option to reset the alignment, to Low, and again the dense cloud in all categories. I redid this for all alignment options, always with the same settings. The dense clouds were exported in local coordinate system (because with the correct UTM projection option the generated clouds were a line of points...I didn't understand this either).
When loading clouds in CloudCompare or Meshlab, only the clouds that aligned to Lowest correspond to the actual coordinates of the control points used, the others are shifted and rotated without consistency.
I put a sample image in the attachment. The purple points are my GCPs, the best fit cloud is the Lowest alignment and the other is Medium alignment.
Can anyone help me understand what is going on? It is the first time I see this kind of problem in 5 years!! I really need this to work because it is part of my PhD
Thanks!
Update: made the same tests - dataset, GCP, options - with Pix4D and the generated dense clouds align perfectly!
redid the clouds in Metashape without the GCP and the alignment problem persists even without references
I'm doing some testing with the latest version of Metashape Pro, and found a problem that I can't understand.
I have a set of 83 images and 6 GCPs, which were manually identified in all of them (more than 20 images per point). I generated the sparse cloud with the Lowest alignment (generic and reference preselection enabled) and the dense cloud in all categories in sequence. Then I generated a new sparse cloud, with the option to reset the alignment, to Low, and again the dense cloud in all categories. I redid this for all alignment options, always with the same settings. The dense clouds were exported in local coordinate system (because with the correct UTM projection option the generated clouds were a line of points...I didn't understand this either).
When loading clouds in CloudCompare or Meshlab, only the clouds that aligned to Lowest correspond to the actual coordinates of the control points used, the others are shifted and rotated without consistency.
I put a sample image in the attachment. The purple points are my GCPs, the best fit cloud is the Lowest alignment and the other is Medium alignment.
Can anyone help me understand what is going on? It is the first time I see this kind of problem in 5 years!! I really need this to work because it is part of my PhD
Thanks!
Update: made the same tests - dataset, GCP, options - with Pix4D and the generated dense clouds align perfectly!
redid the clouds in Metashape without the GCP and the alignment problem persists even without references
