Cannot load information on dwoslam, which means that it is not yet in our index. Please see this page for information on how to submit your repository to our index.

Package Summary

dwoslam is a VSLAM system that is able to track an RGB-D camera, like the Microsoft Kinect, combining visual features and realtime ICP to achieve high performance and accuracy. Then, it looks for connections between key-frames and uses a double-window technique to globally optimize the whole graph.

* Author: Nicola Fioraio, Kurt Konolige, Boris Lau
* License: BSD
* Source: svn https://kforge.ros.org/dwoslam/svn/trunk

Overview

The package is still under development. At the time of writing, it provides an easy-to-use interface to add Generalized-ICP constraints to an existing g2o graph and a KLT+ICP camera tracker to effectively combine those constraints with 2D features in a uniform framework.

Getting Started

First, get the code from kforge.ros.org and rosinstall some dependencies:

mkdir overlays
svn co https://kforge.ros.org/dwoslam/svn/trunk overlays/dwoslam
rosinstall overlays /opt/ros/diamondback overlays/dwoslam/dwoslam.rosinstall

Then, build everything:

source overlays/setup.sh
rosmake --rosdep-install dwoslam

Run The Realtime GICP Demo

To run a simple realtime GICP demo, launch your Kinect (see openni_kinect), or start playing a bagfile; then:

rosrun dwoslam realtime_gicp_demo rgb:=/image_color/topic depth:=/depth/topic info:=/rgb/camera_info/topic

To see a complete list of options:

rosrun dwoslam realtime_gicp_demo --help

The known commands are shown in the terminal and are caught by the OpenCV window:

spacebar

start/pause

r

reset (restart from scratch)

c

close the loop (experimental)

Run The KLT+ICP Tracker

To run the KLT+ICP tracker:

rosrun dwoslam dwo_node [file.bag] --topics /camera/rgb/image_color /camera/depth/image /camera/rgb/camera_info --harris --do-icp

If no bagfile is provided it starts an online session. To see a complete list of options:

rosrun dwoslam dwo_node --help

Commands caught by the OpenCV window:

• online session

  spacebar

  restart

  g

  write the graph to pdf

  testing session

  spacebar

  restart

  g

  write the graph to pdf

  s

  skip frame

  c

  continue (track this frame)

  f

  go to frame no...(write the frame number in the terminal window)

  t

  go to timestamp...(write the timestamp -ns- in the terminal window)

  e

  exit

When the graph is drawn in the viewer, the program could stop: simply press 'q' in the 3D window.

C++ API

The code is still under development, the API is subject to change and therefore the code is not yet publicly available. If you are a developer and want to contribute, please ask to be added to the developer list.