/visual_odometry

Primary LanguageC++

Visual Odometry Based on RGB images

This packages listens to an RGB image topic and calculates the camera motion on a 2D plane based on feature tracking (Lukas-Kanade method) and optical flow.

It requires a calibrated RGB camera with its parameters published on the ROS parameter server. An example file can be found in the config folder.

By default the node listens for images in the /usb_cam/image_raw topic and publishes the estimated motion as a geometry_msgs/Twist message on the /twist/visual

Example Run

user@host:~$ roslaunch visual_odometry usb_cam.launch

Depends

  • ROS Kinetic
  • OpenCV 2 and greater
  • ROS Image Transport

Generate Documentation

user@host:~$ doxygen Doxyfile

TODO

  • Implement some of the camera attribute getters/setters
  • Try different motion tracking methods