We are going to install a single package called perception, which is a meta package of ROS containing all the perception related packages such as OpenCV, PCL, and
so on.

• In ROS Jade

    $ sudo apt-get install ros-jade-perception  

• In ROS Indigo

    $ sudo apt-get install ros-indigo-perception  

The ROS perception stack contains the following ROS packages:

• image-common: This meta package contains common functionalities to handle an image in ROS. The meta package consists of the following list of packages (http://wiki.ros.org/image_common):
  • image_transport: This package helps to compress the image during publishing and subscribes the images to save the band width (http://wiki.ros.org/image_transport). The various compression methods are JPEG/PNG compression and Theora for streaming videos. We can also add custom compression methods to image_transport.
  • camera_calibration_parses: This package contains routine
    to read/write camera calibration parameters from an XML file.
    This package is mainly used by camera drivers for accessing calibration parameters.
  • camera_info_manager: This package consists of routine to save, restore, and load the calibration information. This is mainly used
    by camera drivers.
  • polled_camera: This packages contains interface for requesting images from a polling camera driver (for example, prosilica_camera).

• image-pipeline: This meta package contains packages to process the raw image from the camera driver. The various processing done by this meta package are calibration, distortion removal, stereo vision processing, depth image processing, and so on. The following packages are present in this meta package for this processing (http://wiki.ros.org/image_pipeline):
  • camera_calibration: One of the important tools for relating the 3D world to the 2D camera image is calibration. This package provides tools for doing monocular and stereo image calibration in ROS.
  • image_proc: The nodes in this package act between the camera driver and the vision processing nodes. It can handle the calibration parameters, correct image distortion from the raw image, and convert to color image.
  • depth_image_proc: This package contains nodes and nodelets
    for handling depth image from Kinect and 3D vision sensors.
    The depth image can be processed by these nodelets to produce
    point cloud data.
  • stereo_image_proc: This package has nodes to perform distortion removal for a pair of cameras. It is same as the image_proc package, except that it handles two cameras for stereo vision and for developing point cloud and disparity images.
  • image_rotate: This package contains nodes to rotate the input image.
  • image_view: This is a simple ROS tool for viewing ROS message topic. It can also view stereo and disparity images.
• image-transport-plugins: These are the plugins of ROS image transport for publishing and subscribing the ROS images in different compression levels or different video codec to reduce the bandwidth and latency.
• laser-pipeline: This is a set of packages that can process laser data such as filtering and converting into 3D Cartesian points and assembling points to form a cloud. The laser-pipeline stack contains the following packages:
  • laser_filters: This package contains nodes to filter the noise in the raw laser data, remove the laser points inside the robot footprint, and remove spurious values inside the laser data.
  • laser_geometry: After filtering the laser data, we have to transform the laser ranges and angles into 3D Cartesian coordinates efficiently by taking into account the tilt and skew angle of laser scanner.
  • laser_assembler: This package can assemble the laser scan into a 3D point cloud or 2.5 D scan.
• perception-pcl: This is the stack of PCL-ROS interface.
• vision-opencv: This is the stack of OpenCV-ROS interface.