Differences

This shows you the differences between two versions of the page.

--- teaching:se-kiba:perception-tutorial [2013/04/24 20:32] – balintbe
+++ teaching:se-kiba:perception-tutorial [2016/05/19 09:19] (current) – external edit 127.0.0.1
@@ Line 1: / Line 1: @@
-=====Perception Tutorial=====
+=======Perception Tutorial=======
-More detailed description to be added soon\\
-Code will be available at:\\
+Tutorial describing first steps needed to start processing data from the Kinect, in the form of Point Clouds,  using tools included in ROS Fuerte.
-[[https://github.com/ai-seminar/perception-tutorials.git]]
-==Viewing and recording data from the Kinect==
+==== 1. Prerequisite ====
-  * viewing with //rviz// (point clouds) and //image_view //(images)\\
-  * record using bag files or save clouds in *pcd files using tool in ros_pcl\\
-  * replay bag files, view them in rviz\\
-  * subscribing to the PointCloud topic from C++
-==Processing Point Clouds==
+Follow the steps described in [[software:ros:installation|Recommended procedure for installing ROS]] and [[software:ros:getting-started|Getting Started with ROS]]
-  * reading in a pcd file
+If you are using a PrimeSense device (Xbox Kinect, Asus Xtion) check whether //ros-fuerte-openni-kinect// is installed. You can do this by running the following command in a terminal:
-  * using PCLVisualizer
+  dpkg -l | grep  ros-fuerte-openni-kinect
-  * removing nans
+If not present install using:
-  * filtering based on axes
+  sudo apt-get install ros-fuerte-openni-kinect
-  * downsampeling POint Cloud
+Check out the code presented at seminar in your ros-workspace from:\\
+[[https://github.com/ai-seminar/perception-tutorials.git]] \\
+Prerecorded bag file can be found in ../tutorial_pkg/data/.
+==== 2. Viewing and recording data from the Kinect ===
+== Bag files and Rviz ==
+For a PrimeSense device: connect it to your PC and run:
+  roslaunch openni_launch openni.launch
+Run rviz:
+  rosrun rviz rviz
+Add a new PointCloud2 display type to rviz, and choose // /camera/depth_registered_points // in the topics field. If rviz returns errors, change //Fixed Frame// in //Global Options// from //<Filxed Frame>// to one of the available ones in the dropdown list (e.g.  /camera_link)
+Use //rosbag// to record data in a bag file, e.g.:
+  rosbag record /camera/depth_registered/points /tf
+Note: /tf is needed if you want to view the recorded data using rviz.
+Play back a bag file using:
+  rosbag play filename.bag --loop
+More detail and a more elegant way of saving data from a Kinect to bag files can be found [[http://www.ros.org/wiki/openni_launch/Tutorials/BagRecordingPlayback|here]]
+In order to save Point Clouds to *.pcd files run:
+  rosrun pcl_ros pointcloud_to_pcd /input:=/camera/depth_registered/points
+== Image_view ==
+View rgb image:
+  rosrun image_view image_view image:=/camera/rgb/image_color
+View depth image:
+  rosrun image_view image_view image:=/camera/depth/image
+== From a ROS node ==
+To see how you subscribe to a topic from a ros node take a look at //subscriber.cpp// from the source code.
+==== 3.Processing Point Clouds ====
+The following are presented in //tutorial.cpp//, for more detailed description check the links attached.
+  * reading in a pcd file[[http://pointclouds.org/documentation/tutorials/reading_pcd.php#reading-pcd|link]]
+  * using PCLVisualizer[[http://pointclouds.org/documentation/tutorials/cloud_viewer.php#cloud-viewer|link]]
+  * removing nans
+  * calcultaing normals[[http://pointclouds.org/documentation/tutorials/normal_estimation.php#normal-estimation|link]]
+  * filtering based on axes[[http://pointclouds.org/documentation/tutorials/passthrough.php#passthrough|link]]
+  * downsampling the Point Cloud[[http://pointclouds.org/documentation/tutorials/voxel_grid.php#voxelgrid|link]]
+  * RANSAC plane fitting, and extracting indices [[http://pointclouds.org/documentation/tutorials/planar_segmentation.php#planar-segmentation|link1]][[http://pointclouds.org/documentation/tutorials/extract_indices.php#extract-indices|link2]]
+For further questions contact: //**balintbe** at **tzi** dot **de**//