FndrPlayer39's profile - activity

Does anybody know how I can get a hold of Alexander Shishkov? It looks like he wrote the EPnP algorithm code and is the most equipped to answer my question. Thank you!

Have you tried to reproject the image points to see what is happening visually? Maybe there is an angle "wrapping" somewhere. For example, if you have a vehicle that is heading approximately 0 degrees (North), you might have a few readings at 359, 0, 1 degrees.

I also know that the rotation vector (rvec) is expressed as "compact" representation of Rodrigues parameters, which are not easy to visualize. You will want to use cv::Rodrigues function to convert your rvec to easier to visualize Euler angles (roll, pitch, heading).

The iterative methods used for pose estimation require an initial guess. It is possible that the method will converge to an incorrect pose if given a bad initial guess. I would recommend using the EPnP algorithm which is non-iterative.

I am using OpenCV's EPnP algorithm to solve for the pose of an aircraft with a known pattern of LEDs. I have been able to get pose estimation working without too much trouble, but I am now in the writing phase of my thesis and have some questions about how the algorithm actually works. I was hoping that someone could point me towards some information that might help to put me on the right track.

I have found this paper (http://cvlabwww.epfl.ch/~lepetit/papers/lepetit_ijcv08.pdf) to be very helpful in understanding how the four control points are found in the camera frame of reference, but the paper does not go on to explain how the rotation-translation matrix [R|t] is actually found, which gives us the pose information.

I have looked through the OpenCV source code and have located the "compute_R_and_T()" function, but have struggled to understand what all is going on. Are there any technical papers or OpenCV documentation that could help me out?

Thanks,

Mike

Is there any documentation or references that can explain how the joint rotation-translation matrix [R|t] is computed once the position of the 3-D points are found in camera coordinates using the EPnP algorithm? I have found the EPnP paper (http://cvlabwww.epfl.ch/~lepetit/papers/lepetit_ijcv08.pdf) to be helpful in understanding how the 3-D locations are found in camera coordinates, but the paper does not explain how [R|t] is determined.

Thank you to anyone who can point me in the right direction. I am working on my Master's thesis and need to understand how [R|t] is found so that I can write about it.