I am trying to understand a paper on visual odometry and I am having troubles understanding how a map is generated using a stereo camera as sensor.

The paper I am reading is SVO: Semi-Direct Visual Odometry for Monocular and Multi-Camera Systems, and at pag. 5, it starts describing how to build a map using a stereo camera.

I have undestood that to build a map the concept of disparity is used and the difference from the disparity is projected along the epipolar line to have an idea of the distance. But in the paper it says that it is used a recursive Bayesian depth filter to estimate depth.

How does the recursive Bayesian depth filter is used in this context is what I don't understand. I am self learning, so I don't have a lecture where I have studied this to refer to.

Can somebody please help me understand?

  • $\begingroup$ A couple of things. First you should probably pick a different VO system than SVO(I recommend VINS Fusion/Mono). SVO does a couple of weird tricks that probably add to your confusion. This being a perfect example. Second, you should start by understanding the original SVO paper. Rather then jumping directly into the improved version. Finally you have asked a lot of questions regarding the same paper. To me it seems like you lack a lot of the prequisite knowledge needed to understand this paper. So I advise you to study up on the basics, before tackling something as complex as SVO. $\endgroup$ – edwinem Apr 17 at 18:56

Your Answer

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

Browse other questions tagged or ask your own question.