This package is meant to be a simple stub for use in the NIST Solutions in Perception contest.

Install

Prerequisites

First, make sure that you have installed the Diamondback release of ROS

• http://www.ros.org/wiki/diamondback/Installation/Ubuntu

Then, verify that you have rosinstall :

• http://www.ros.org/wiki/rosinstall

Also please make sure that you install:

• The openni kinect drivers - http://www.ros.org/wiki/openni_kinect

sudo apt-get install ros-diamondback-openni-kinect

• pcl_visualization - http://www.ros.org/wiki/perception_pcl_addons

sudo apt-get install ros-diamondback-perception-pcl-addons

install tod_stub

Now, open a terminal and create a directory to work in. We'll use rosinstall to setup a directory that has all the dependencies. The following will install everything you need to get started in you.

The following line will install everything in under the path: ~/tod_stub_dev

rosinstall ~/tod_stub_dev https://code.ros.org/svn/wg-ros-pkg/branches/trunk_diamondback/stacks/object_recognition/tod_stub/tod_stub.install

• You must then source setup.bash in ~/tod_stub_dev either "by hand" or put source ~/tod_stub_dev/setup.bash in your .bashrc file.

That just installed:

• tod_stub - The code you'll be working with directly.

• tod_core - Core Textured Object Recognition library.

• tod_training - Textured Object Training library.

• vision_bleeding - misc vision algorithms

3

For CS324 Class, use of fiducial tracking:

"Fiducial" will become your perception code stub as it connects with OpenCV, PCL and the kinect camera. So:

• roscd fiducial

Print out the calibration fiducial pattern in data/5x3.svg

Check that everything runs:

1. plug in a kinect camera
2. Open up 4 terminals (or 4 tabs in one terminal)

3. In the first, run roscore

4. In the second, run roslaunch openni_camera openni_node.launch 

5. In the third, go to the data directory roscd fiducial/data

   • We go to this directory because that's where the fiducial.yml file is.

     • Otherwise, you may append the argument --fiducial data/fiducial.yml

   • Then: rosrun fiducial pose_estimator camera:=/camera/rgb

6. In the forth window, to view all this: rosrun rviz rviz

   • Set rviz settings as follows:

7. 

8. You can run rosrun fiducial pose_estimator camera:=/camera/rgb  with other options (append --help to get these options):

   • Append --use_3d To use 3D information for pose fitting instead of OpenCV's solvePNP().

     • 3D information is more accurate since small differences in 2D estimation of points can cause large pose differences.

   • Append --verbose To show an OpenCV window tracking the chessboard

     • Both options may be used together