I’m doing a photogrammery project to create models and maps in west texas(so not much uniqueness to the images to use lots of control points. I did my initial batch of 600 with”use camera priors” On, got good alignment with minimal residuals(tails). If i bulk add the remaining 600pics(which do have some overlap with the exsisting model on one side, it rotates the whole new set of images, leaving enormous residuals as well as spinning the added image (which is aligned with itself) against the first image about 30deg. If i try to align all images with priors off i get 2 seperate models. I did try to align the first batch with priors on then lock the positions, then add a fe of batch two at a time and that came out wonky with parts aligned and parts tring to attach about 30 deg off. Any ideas to make this mash together?
Set 1 that works with priors Set 2 that gets off axis with camera priors Attempt to lock cameras on set 1 and add set 2 incrementally
Hi Tim, you can check this post: Banana Effect - What To Do If My Model is Bent | Tutorial
It is not the same problem, but there are some solutions which could work also for your case.
I suppose the best option for you is also the GCP usage (but probably those weren’t measured). In such case, if you don’t need the correct position or scale, you can use the computation without the camera prior poses and try to add all images into one project and align it together.
There is also an option that you will need to use more control points. As I can see some, but those look to be placed only on one component/dataset.