RTAB-Map loop closure issue

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

RTAB-Map loop closure issue

ammar.n.abbas
during mapping, while driving forward and reverse to capture similar point of view, the loop closure point detection is crossing each other.

Loop closure hypothesis rejected 

I have used an optical transform for all cameras

'''
<arg name="pi/2" value="1.5707963267948966" />
<node pkg="tf" type="static_transform_publisher" name="base_link" args="0 0 2 -$(arg pi/2) 0 -$(arg pi/2) ego_vehicle ego_vehicle/depth_front 0" />
'''

tf frames
Reply | Threaded
Open this post in threaded view
|

Re: RTAB-Map loop closure issue

matlabbe
Administrator
Do you mean the crossing blue lines on the left view between images of the loop closure? It means that many features have been wrongly matched between the frames. I would be more concerned on the warning messages on the terminal to know why the loop was rejected.
Reply | Threaded
Open this post in threaded view
|

Re: RTAB-Map loop closure issue

ammar.n.abbas
Yes, the blue lines. At some points, the loop hypothesis was accepted but still, the lines were not parallel and crossing each other.
Reply | Threaded
Open this post in threaded view
|

Re: RTAB-Map loop closure issue

matlabbe
Administrator
In the main window, lines represent features with same word ID, not the actual inliers used for transformation estimation. So some lines can cross between features not representing the same real feature, but those between having same word (descriptor).
Reply | Threaded
Open this post in threaded view
|

Re: RTAB-Map loop closure issue

ammar.n.abbas
Oh okay, thank you so much...