Turn a point cloud into an image with Python

asked 2011-07-06 12:28:01 -0500

Poofjunior
88 ●7 ●10 ●15

updated 2014-01-28 17:09:59 -0500

ngrennan
1 ●1 ●1

Hi, everyone,

Recently, I tried to run through the pcl_ros tutorial to convert a point cloud from a Kinect to an image. My hope is to be able to do the exact same thing in Python, not C++, and then work with the image using the OpenCV Python interface.

here's the tutorial link: http://www.ros.org/wiki/pcl_ros/Tutor...

I hit a small snag, though. It seems like the tutorial is referencing a Ros Topic from openni_camera called /camera/depth/points2, but I can't seem to find this topic in list of topics published by openni_camera.

Rather, I found: /camera/depth/points, but this doesn't seem to generate an image when I make this change in the tutorial.

Will getting the image from /camera/depth/points2 be necessary to fulfill my overall goal?

Also, are there any other ways to easily convert a Point Cloud from the Kinect to an image in Python? Many Thanks in advance for everyone's words of wisdom!

edit retag flag offensive close merge delete

add a comment

0

answered 2011-07-06 17:57:25 -0500

fergs

13902 ●59 ●157 ●196 http://www.robotandchisel.com

There are currently no PCL-python bindings, so there is no single function to convert a point cloud into an image. You can however subscribe to /camera/depth/image, which is already an image and has 32-bit float valued pixels (values are in meters). It can be converted easily into a cvMat using cv_bridge (see this post for further details).