OpenCV and ROS

The purpose of this lab is to gain experience incorporating vision code into ROS nodes and making use of depth data from the Kinect.

Resources

• cv_bridge Python Tutorial
• cv_bridge Python API

Exercises

• Set up your workspace:

  $ cd cs354_ws
  $ mkdir src
  $ cd src
  $ catkin_init_workspace
  

• Download and unzip three_d_vision.zip into the src folder. This package has the following contents:

  three_d_vision/
      package.xml
      CMakeLists.txt
      src/                         <-- Importable Python modules go here.
          three_d_vision/          <-- Package for importable modules.
              __init__.py          <-- This file turns the folder into a package.
              point_cloud2.py      <-- Module for reading PointCloud2 messages.
              kinect_subscriber.py <-- Module for subscribing to BOTH 
                                       PointCloud2 and image messages
                                       simultaneously.
      rviz/                        <-- rviz configuration files.
          vision.rviz
          depth.rviz
      launch/                      <-- Launch files
          view_vision.launch
          view_depth.launch
      setup.py
      scripts/                     <-- Scripts that create Python ROS nodes. 
          kinect_vision_demo.py
          kinect_depth_demo.py
  
  

• Execute catkin_make in the root folder of the workspace. This is necessary to properly set up the Python package for imports. At this point, you should either open a new terminal or source the setup.bash script in your devel directory.
• Read over the code in scripts/kinect_vision_demo.py .
• Launch view_vision.launch and look at the published image topic. You can look at the topic in rviz, or by starting an image_view node:

    $ rosrun image_view image_view image:=/red_marked_image
  

• Read over the code in scripts/kinect_depth_demo.py .
• Launch view_depth.launch and look at the published topics in rviz. You should be able to see the location in space of the red pixel relative to the robot.