Query regarding Kinfu in offline mode

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

Query regarding Kinfu in offline mode

Hassan AFZAL
Hello,

I am trying to use the KinFu implementation in PCL in offline mode with data captured from Asus Xtion Pro Live sensor. I think the question is naive but it would be great if you can help me. After running KinFu I extract the point cloud from tsdf volume. Now when i plot this cloud together with one of the frames of input data, it seems like the extracted point cloud is shifted by some transformation. Can you please tell me what to do to make the extracted data in the same coordinate system as the input data?

Best regards,
Hassan
_______________________________________________
[hidden email] / http://pointclouds.org
http://pointclouds.org/mailman/listinfo/pcl-users
Reply | Threaded
Open this post in threaded view
|

Re: Query regarding Kinfu in offline mode

kwaegel
Administrator
Hassan AFZAL wrote
I am trying to use the KinFu implementation in PCL in offline mode with data captured from Asus Xtion Pro Live sensor. I think the question is naive but it would be great if you can help me. After running KinFu I extract the point cloud from tsdf volume. Now when i plot this cloud together with one of the frames of input data, it seems like the extracted point cloud is shifted by some transformation. Can you please tell me what to do to make the extracted data in the same coordinate system as the input data?
Well, the point cloud from each depth image is in coordinates relative to the camera (i.e. the camera is always the origin). The KinFu cloud, however, is in it's own separate coordinate system, which is initialized to be the same as the first camera frame.

If you want to get each depth frame in KinFu coordinates, I think you need to save the camera pose ( _kinfu.getCameraPose(..) ) after integrating each frame and use that to transform the desired frame data into Kinfu space.

(You could also use the inverse and transform the Kinfu cloud into camera space, but it still depends on the relative camera location for each frame.)