In summer 2012, I refined my Kinect calibration ROS package in preparation for a public release. I also played around a little with the Point Cloud Library and started working on making the turtlebot_block_manipulation demo work on PR Lite. I posted my summer 2012 updates online because the Interaction Lab had many student workers during that time.
Summer 2012 milestones:
- Essentially finish the Kinect calibration project. This means:
- Add a BSD license, first emailing the upstream developers to find out what they think of using their code. (done 7/10/2012)
- Improve how the program stores calibrations in the yaml file. (done 6/9/2012)
- Convert the shell scripts such as calibrate_2_kinects.sh into launch files (for Interaction Lab use only). (done 6/9/2012)
- Update the readme?
- Put the multi_calibrate package in a public repository and add a web page at ros.org. (draft wiki page here)
- Use the Point Cloud Library to combine 2 or more point clouds and downsample them.
- Combine person tracking using the Kinect with leg tracking using a lidar. This means:
- Do some basic research on the relevant packages and make a list of milestones. :)
- Complete the milestones I come up with for (1).
I finally updated the very outdated description of my Kinect calibration project on this web page.
8/4/2012, 8/11/2012, 8/18/2012
I worked on making the turtlebot_block_manipulation demo work on PR Lite, one of the proposed alternatives/additions to the third milestone. (This isn't as simple as it sounds due to PR Lite's more complex body.)
I added a local readme in readme.txt based on the wiki, and renamed the old readme to readme_old.txt.
I added BSD licensing headers to the multi_calibrate package as suggested by a maintainer of the package it uses code from. Note that according to camera_pose_calibration's changeset history, calibration_transform_publisher.py was never modified by Vijay Pradeep, so I didn't include Vijay in the authors list for that file.
7/7/2012 - 7/8/2012
I was able to downsample (but not yet combine) a Kinect point cloud in a new package called pcl_test_andrew in the Interaction Lab svn repository. I also discussed alternates to the third project for this summer that would be applicable to PR Lite as well as the Interaction Lab. The ideas we came up with were making the turtlebot_block_manipulation demo work on PR Lite and combining (or at least transforming to the same tf frame) the data from PR Lite's Kinect and tilting lidar.
I finally finished a draft of the wiki page for the package, and sent an email to Ross about my remaining questions.
I made new launch files in the multi_calibrate_interaction package to replace the calibrate_X_kinects.sh shell scripts, added error checking to the yaml file code, and updated some documentation.
5/21/2012 - 5/25/2012
On Wednesday, I wrote an initial attempt at storing the Kinect calibrations in a yaml file instead of a bag file. On Friday, I improved how the code transforms calibrations to each other so that it doesn't store duplicates of camera entries. I want to test this latest code myself a little more, and after that it should be tested with several Kinects in the Interaction Lab since I only have 2 Kinects at home.