Reading this paper on visual odometry, where they have used a bearing vector to parameterize the features. I am having a hard time understanding what the state propagation equation for the bearing vector term means :

Bearing vector and inverse depth parametrization

The vector N is not mentioned in the equations, so its not very clear what it does. Would really appreciate if someone would help me understand it :)


The text says:

[...] where $N^T (\mu)$ linearly projects a 3D vector onto the 2D tangent space around the bearing vector $\mu$ [...]

  • $\begingroup$ Yes, but I am not able to get an intuitive understanding of what that means, plus how the equation comes up with those three terms, i.e. relating bearing vector with w and v and w(u,i) $\endgroup$
    – rookie
    Mar 4 '16 at 14:53
  • 2
    $\begingroup$ Maybe you could rephrase your question accordingly. $\endgroup$
    – Jakob
    Mar 8 '16 at 9:42

It appears to me that the notation represents taking an arbitrary 3-vector, applying a rotation matrix to the 3-vector based on the bearing 3-vector, and then removing one of the dimensions of the result. If the arbitrary 3-vector and the bearing 3-vector are the same, then the result of the rotation should be [ 1 0 0 ], and the first coordinate would be removed, since the projection of that vector onto the plane orthogonal to it is zero. For an arbitrary 3-vector, you get the components of the projection onto a plane that is normal to the bearing 3-vector.

As to how to compute the rotation matrix based on the bearing 3-vector, I would defer to the following:



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.