Stereo solvePnP

, ,
1

I have a robot moving around a static environment with AprilTags in known locations. The current system I have, uses solvePnP on a single camera to get the location of the robot from the AprilTags, but due to tag ambiguity, noise and motion blur this isn’t always accurate.

I have the ability to get a stereo camera which I was hoping would help some of these issues, but looking through the solvePnP docs, I don’t really know how to use data from multiple cameras.

Does anyone know where to start with this?
Do I somehow use stereoRectify first before solvePnP?
Do I use solvePnP separately and then somehow merge the resulting matrices?
Is there a better way?
Will switching to stereo camera really help the issue?

2

Stereo will get you depth. It sounds like you are just trying to get better results from solvePnP. I don’t have any experience with AprilTags, but if they have the same issues as Arucos they might not be a very good at providing stable features. Can you post some images, including the AprilTag detection (drawn to the image).

Also if you are having issues because of motion blur, stereo won’t help that.