OAK-D misalignment effects in Rtabmap

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

OAK-D misalignment effects in Rtabmap

FPSychoric
Hi I got a OAK-D camera which have a sensor misalignment which Luxonis think is normal, but I don't think so.

https://youtu.be/HyfL8lXYvUo


It will make permanent band (top half right) with false distance close to 0, or at least much less than the real distance, not the red band in the blue area.

How you think this would affect to your software, will it make false map areas, areas that cannot be cleaned by the nav stack, loop closure or VIO damage?

Thanks, I cannot test it in a robot due the bad development state of OAK-D in ROS and I never had such issue in a commercial stereo camera.

How you think it would affect to a drone like this in VIO or Slam?  

Reply | Threaded
Open this post in threaded view
|

Re: OAK-D misalignment effects in Rtabmap

matlabbe
Administrator
Hi,

In the video, if between 0:04 and 0:08 you are switching between left and right cameras, the stereo rectification is wrong as pixels are also shifting vertically (look at the box on top right).

Which launch are you using to bring up the oak-d? I recommend using stereo_inertial_node.launch instead of the stereo depth ai example with which I had a similar issue.

cheers,
Mathieu