Triangulation of rectified image points in multiple views

Question

I work with a set of calibrated images that form a ring around a foreground object ( 1 ). I used the Fusiello ( 1 ) method to fix adjacent pairs of images, and then I performed a mismatch assessment.

When I take coincident points from a stereo pair and triangulate them, it forms an exact point cloud. Unfortunately, when I triangulate points from another pair of stereo images, this point cloud never aligns correctly with the original cloud.

Should calibrated point clouds with corrected images merge automatically?

Thanks in advance for any help you can offer.

Answer 1

This may be due to the accuracy of the calibration - both internal (for example, the same camera model, and how it copes with distortions), and external (i.e. the camera poses in real space). Together, of course, this dictates the maximum accuracy of your projection.

Do you have an indicator of camera calibration error - in terms of MSE redesign?

The cumulative error often manifests itself in my experience, if you simply repeat the subsequent images. Some form of global optimization often needs to be performed for the first correct positions for all camera poses.

. , , / ( /), , ( - / / ..).

, "" ? ( ICP ...). "" ? , ?

, (- ) , "" . ( → → → ...)

MeshLab BTW?