Translation of point cloud from multiple D435i devices
I have 2 D435i devices mounted 90 degrees to each other. I would like to 'translate' the point cloud from one sensor to the same coordinates of the other sensor.
I am a technically competent developer, however my head never fully grasped the 3D world and the math required to fully understand an answer of 'apply a translation matrix' to the second one to match the first one.
Can someone offer a suggestion as to what I can read/explore to write the code to do this translation. I am not trying to get this perfect, I am not trying to reconstruct complicated scenes, I am dealing with simple geometric structures and only need 'close' estimates - since sensors are at 90 degrees, there is no overlap between the two images.
Thanks for your time and any input,
Ricky
-
Your scenario sounds similar to a tutorial published by Intel for combining the point clouds of two cameras facing the same scene but at 90 degree differences. In this tutorial, the vision softwares ROS and Rviz are used to generate the point cloud image, reducing the amount of programming that you would need to do,
https://github.com/IntelRealSense/realsense-ros/wiki/Showcase-of-using-2-cameras
-
I had previously read through that article and dug a bit into the code, but it is based on ROS and has a lot of baggage which was getting hard to follow. For R&D. ROS is okay, but for production, it is not a good solution. I will dig a bit deeper. It seems that device translation is so common that there would be solid tutorial or example for doing this.
Thanks for the time/reply.
-
If you feel confident in your developer skills, a multi camera point cloud stitching system that is complex is the one developed by the CONIX Research Center at Carnegie Mellon.
https://github.com/conix-center/pointcloud_stitching
Please sign in to leave a comment.
Comments
3 comments