CODEWITHSUNDEEP
opencv3d-reconstructionmultiviewpose-estimation
batuman
batuman

Reputation: 7304

Relative pose optimization using 5-point algorithm and RANSAC

I am trying to get 3D reconstruction from uncalibrated multi-view images. I don't know the intrinsic parameters of the camera

I have SIFT features.

What I like to do is filtering out-liers using the 5-point algorithm in combination with RANSAC, so that I can proceed for the relative pose optimization and triangulation of the points matched.

Opencv has one API findEssentialMat(); That API needs focal and pp. Where I can have focal and pp? Is this API findEssentialMat() the right one I have to use for the pose estimation? If my approach is wrong, is there any API closer to what I want to achieve in OpenCV? Thanks

Upvotes: 2

Views: 1366

Answers (1)

user3146587
user3146587

Reputation: 4320

I don't know the intrinsic parameters of the camera. [...]

[...] proceed [with] the relative pose optimization and triangulation of the points matched.

Where I can have focal and pp?

Then findEssentialMat() can not be used in this situation as it requires the intrinsic parameters, which are the focal length and principal point (focal and pp arguments).

First calibrate the camera to recover these parameters. Then pose estimation and 3D triangulation will be possible using OpenCV functions.

Upvotes: 1

Related Questions