srv/viso2

how to apply on carla-ros-bridge

Opened this issue · 5 comments

Hello I'm struggling to use libviso with my available topics. I want to use stereo_odometry on the right and left camera which coordinate systems are already parallel and only a translation in y in vehicle coordinates is applied. However after 'catkin_make' I neither find rosun stereo_odometer nor any of the other proposed packages. However, when I do 'catkin_make' they seem to be build.

image

image

Hi, here is a demo.launch of how to use stereo_odometer, hope it is useful.
(https://github.com/srv/viso2/blob/25db160641faa5a1cb09147716d15dbba86c6bd5/viso2_ros/launch/demo.launch)

I was able to use it and it starts now but I have an instant segmentation fault
image

I used a bag file which was downloaded from here, and didn't encounter segmentation fault, maybe you can have a try.

I found out that if I add some roll to one of the cameras it's not compensated in image_rect by stereo_image_proc. Probably the base_link isn't established correctly.
image
that's the transformation done by rosrun tf tf_echo ego_vehicle ego_vehicle/camera/rgb/left
image
But I think I get my segmentation fault because the tf couldn't be done and is to contrary to the identity matrix. left or right frame is no difference
image

Pretty sure it is the same issue as this one #63 (comment)

as @ysong-aceinna mentions, using the height and width of the bag-files images will most likely fix the issue.