use kinect-like frame settings

[k-okada]

There are several atttempt to support 'kinect-like' settings, https://github.com/ipa320/softkinetic/pull/44, https://github.com/ipa320/softkinetic/pull/58

I think the definition of kinect-like consists of

• publish point cloud in 'depth frame'
• point cloud coordinates is : 'kinect-like' coordinates cloud means +z for forward +y is down and +x for right-ward
• use rgdb_launch/launch/kinect_frames.launch style to set camera_link to depth/rgb optical frame

Here is the examples from openni2.launch

https://github.com/ipa320/softkinetic/pull/63/commits/159419179d0312b40dba932c9bb3cd9d4963f7ad includes

 +  <node pkg="tf" type="static_transform_publisher" name="softkinect_rgb_frame_tf"
 +        args="0 0.00090 0  0 0 0 /softkinetic_camera_link /softkinetic_camera_rgb_frame 40" />
 +  <node pkg="tf" type="static_transform_publisher" name="softkinect_depth_frame_tf"
 +        args="0 -0.0090 0.005 0 0 0 /softkinetic_camera_link /softkinetic_camera_depth_frame 40" />

which indicates the translation between rgb and depth frames is '0.00180 0 0.005' which may differ from the urdf settings (0.0250 displacement), I'm not sure if this parameter is general one, but at least for my camera, it improve the depth-camera registration

0.0250 displacement (see left bottom camera iamge which overay point cloud into camera image)

0.00180 0 0.005 displacement

• softkinetic.rviz : update rviz config for kinect-like frame settings
• softkinetic_camera_demo.launch : publish 'kinect_frame' like tf using static_tf_publihser, see rgdb_launch/launch/kinect_frames.launch, maybe we need to remove these frames from urdf?
• 'kinect-like' coordinates cloud means +z for forward +y is down and +x for right-ward
• use depth_optical_frame for cloud header

[gavanderhoorn]

Is this related / a duplicate of #58?

[k-okada]

Yes, this is fixed version of #58, I updated the first comment with rviz images, any comments are welcome.

[gavanderhoorn]

Ah, ok.

You probably want to check that kinetic_dev doesn't already contain these changes.

[k-okada]

this PR conflicts with https://github.com/ipa320/softkinetic/commits/kinetic_dev which includes #58, I think #58 did not support following two points, which I think definition of 'kinect-like' style

• publish point cloud in 'depth frame'
• point cloud coordinates is : 'kinect-like' coordinates cloud means +z for forward +y is down and +x for right-ward

[knorth55]

@k-okada is the camera frame of soft kinetic converted to third one? is this kinect-like?

[k-okada]

As far as I understand , point cloud should have

• +z for forward +y is down and +x for right-ward Which is typical Computer Graphics coordinates systems and use tf to rotate them for robotics coordinate systems which +z is upward

[knorth55]

ok, thank you.

[knorth55]

:+1: I'm for this PR. Sorry for my misunderstanding about kinect-like camera frame.