ROS DAVIS Tracker

You first need to install davis_ros_driver from https://github.com/uzh-rpg/rpg_dvs_ros

ROS workspace

You need a ROS workspace. If you followed the instructions above, this workspace is catkin_ws.

You need to source the setup file:

cd catkin_ws
source devel/setup.bash

Building

catkin build track_init tracker

Running

Before running you need to have the davis camera calibrated with no tangential distortion http://wiki.ros.org/camera_calibration/Tutorials/MonocularCalibration

Easy way

    roslaunch tracker track.launch

or a nodelet version

    roslaunch tracker track_nodelet.launch

Then start tracking with rostopic pub /reset std_msgs/Bool "data: true" -1, you can also use it to reset the tracker (starts from the last pose received) in case it gets lost.

Those are running 5 nodes: davis_ros_driver, track_init, tracker and a couple image_view for visualization. You can run them individually, check the launch files for more details.

With a ros bag

This is useful for debugging or slow step-by-step visualization. If you have a bag data with:

• /dvs/camera_info : sensor_msgs/CameraInfo
• /dvs/events : dvs_msgs/EventArray
• /track/init_pose : geometry_msgs/PoseStamped
• /track/init_rendering : sensor_msgs/Image (optional)

start ros

roscore

and the tracker

rosrun tracker tracker camera_info:=/dvs/camera_info camera_pose:=/track/init_pose events:=/dvs/events

play the bag in pause mode

rosbag play track_data.bag --pause

press s multiple time in the rosbag terminal until the tracker gets the camera_info and camera_pose send a reset event to the tracker to start it

rostopic pub /reset std_msgs/Bool "data: true" -1

return to the rosbag and keep pressing s (step by step) or space (play/pause)

You can also visualize the track_init square detection and the projected map with events:

rosrun image_view image_view image:=/track/init_rendering
rosrun image_view image_view image:=/map_events

ROS Nodes

track_init

Computes a camera pose (geometry_msgs/PoseStamped) from an image feed (known map)

• Publications:
  • pose [geometry_msgs/PoseStamped]: computed camera pose relative to the map
  • rendering [sensor_msgs/Image]: visualization of found map + reprojected map vertices
• Subscriptions:
  • camera_info [sensor_msgs/CameraInfo]: camera parameters
  • image [sensor_msgs/Image]: raw image feed

tracker

Tracks a segment map with events from the camera

• Publications:
  • /map_events [sensor_msgs/Image]: visualization of the tracked map with events (used in red)
  • /tracked_pose [geometry_msgs/PoseStamped]: estimated camera pose
• Subscriptions:
  • /camera_info [sensor_msgs::CameraInfo]: camera parameters
  • /camera_pose [geometry_msgs::PoseStamped]: first camera pose (usually from track_init)
  • /events [dvs_msgs::EventArray]: camera events
  • /reset [std_msgs::Bool]: start&reset flag channel, sending a msgs starts tracking or resets it

Files

├── README.md
├── tracker
│   ├── CMakeLists.txt
│   ├── include
│   │   └── tracker
│   │       ├── efk.h
│   │       ├── slam_line.h
│   │       ├── tracker.h
│   │       ├── tracker_map.h
│   │       └── tracker_nodelet.h
│   ├── launch
│   │   ├── track.launch
│   │   └── track_nodelet.launch
│   ├── package.xml
│   ├── src
│   │   ├── efk.cpp
│   │   ├── slam_line.cpp
│   │   ├── tracker.cpp
│   │   ├── tracker_map.cpp
│   │   ├── tracker_node.cpp
│   │   └── tracker_nodelet.cpp
│   ├── test
│   │   └── test.cpp
│   └── tracker_nodelet.xml
└── track_init
    ├── CMakeLists.txt
    ├── include
    │   └── track_init
    │       └── track_init.h
    ├── launch
    │   └── track_init.launch
    ├── package.xml
    └── src
        ├── track_init.cpp
        ├── track_init_node.cpp
        └── track_test.cpp