openni2_camera

ROS wrapper for openni 2.2 and NiTE 2.2

Note: openni2_camera supports xtion devices, but not kinects. For using a kinect with ROS, try the freenect stack: http://www.ros.org/wiki/freenect_stack

This package derives from https://github.com/OpenNI/OpenNI2 and includes NiTE 2.2 to perform:

• hand tracking and gesture recognition (waving and clicking)
• person detection and tracking
• skeleton tracking

Requirements

Linux Ubuntu 12.04 32-bit with ROS hydro

• OpenNI-Linux-x86-2.2 (version 2.2.0.33)
• NiTE-Linux-x86-2.2 (version 2.2.0.10)
• pal_msgs
• openni2_launch

Installation

OpenNI-2.2

Right now the following files are alread included so you may skip the following steps:

From OpenNI-Linux-x86-2.2.0.33.tar.zip:

Copy Redist/libOpenNI2.so to openni2_camera/lib
Copy Redist/OpenNI2 to openni2_camera/lib
Copy Include/Linux-x86 to openni2_camera/include/OpenNI-2
Copy Include/Driver to openni2_camera/include/OpenNI-2

NiTE-2.2

Right now the following files are alread included so you may skip the following steps:

From NiTE-Linux-x86-2.2.tar.zip:

Copy libNiTE2.so to openni2_camera/lib
Copy Include/* to openni2_camera/include/NiTE-2

pal_msgs

Clone the following repository in your catkin workspace:

git clone [email protected]:pal-robotics/pal_msgs.git
cd pal_msgs
git checkout hydro-devel

openni2_launch

Clone the following repository in your catkin workspace:

git clone https://github.com/ros-drivers/openni2_launch.git
cd openni2_launch
git checkout hydro-devel

Guidelines

How to launch the camera node

roslaunch openni2_launch openni2.launch

How to visualize the rgb image

rosrun image_view image_view image:=/camera/rgb/image_raw

How to visualize the depth image

rosrun image_view image_view image:=/camera/depth/image

How to test gesture recognition

rostopic echo /camera/gestures

Wave in front of the camera or do the 'click' movement, i.e. point forward and bring back the arm

How to visualize the user map

The detected and tracked persons can be visualized in a segmented depth image as follows

rosrun image_view image_view image:=/camera/user_map

How to get tracked persons position

The topic /camera/users of type pal_detection_msgs::PersonDetections contain the 3D position of the tracked persons wrt the depth camera frame

rostopic echo /camera/users

---
header: 
  seq: 675
  stamp: 
    secs: 1401778685
    nsecs: 563214248
  frame_id: ''
persons: 
  - 
    full_body: 
      x: 0
      y: 0
      width: 0
      height: 0
    face: 
      x: 0
      y: 0
      width: 0
      height: 0
      eyesLocated: False
      leftEyeX: 0
      leftEyeY: 0
      rightEyeX: 0
      rightEyeY: 0
      position: 
        x: 0.0
        y: 0.0
        z: 0.0
      name: ''
      confidence: 0.0
    legs: 
      header: 
        seq: 0
        stamp: 
          secs: 0
          nsecs: 0
        frame_id: ''
      position3D: []
      laser_pose: 
        header: 
          seq: 0
          stamp: 
            secs: 0
            nsecs: 0
          frame_id: ''
        child_frame_id: ''
        transform: 
          translation: 
            x: 0.0
            y: 0.0
            z: 0.0
          rotation: 
            x: 0.0
            y: 0.0
            z: 0.0
            w: 0.0
    position3D: 
      header: 
        seq: 0
        stamp: 
          secs: 0
          nsecs: 0
        frame_id: camera_depth_optical_frame
      point: 
        x: 380.052642822
        y: -136.789474487
        z: 801.789489746
---

As it can be seen, the useful information appears in the persons.position3D field