0
$\begingroup$

I am using typical setup with camera and IMU as found on a smartphone. Using this, I would like to estimate the relative pose between the camera and an object (for simplicity, let's say a cube). The series of measurements contain motion around the object, so a monocular visual-inertial dataset.

Now, related to this usecase I have two questions:

  1. Which state-of-the-art algorithms would commonly be used to solve such a problem?
  2. What's a realistic accuracy that could be achieved? (just a rule-of-thumb)
Improve this question
$\endgroup$
1
  • $\begingroup$ When you say moving, is it an actual moving robot or a static platform with moving parts (like a manipulator)? The actual transformation from the object to the camera (static or not) is done using computer vision. Depending on Camera resolution, technic and process time you get different approximations from few centimeters to millimeters. What changes here is if you know the exact position of the camera in relation to the world or not. $\endgroup$
    – Felipe Henrique
    Apr 29, 2019 at 12:47

1 Answer 1

Reset to default
0
$\begingroup$
  1. You need a SLAM to track the object location.

There are some available resources on visual-inertial SLAM. To name some of them, okvis, svo2, orb-slam2 and so on.

  1. If you hire a good slam, error will be less than 10cm.
Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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