-
Notifications
You must be signed in to change notification settings - Fork 216
Description
This is not an issue but rather a question.
I would like to use SfM with a calibrated camera - I looked at the source code a bit. I see there is a "bundle adjustment" done probably in all SfM libraries. Is it possible to plug in (=hard code) my camera parameters (focal length, principal point or maybe even some more?) to your code and get better results?
What I am looking for a classical "multiview" SfM where one has 100 photos of a single object but I need to reconstruct from continuous sequence (moving car) - I want to reconstruct from 2 consecutive frames (or maybe more, but the fewer the better) of a monocular sequence. I can imagine this can be done with bundle adjustment too but I guess when I know both my camera parameters and the exact camera translation/rotation then 3D reconstruction may be done without bungle adjustment (much simpler) and maybe even more accurately?
I know my absolute translation and want to employ OpenCV's recoverPose and then do very simple triangulation to get my 3D scene. Could you give me some hints please?