|
Project Information
Featured
Links
|
What's newMay 2012 - Five new videos (720p) from other community data sets can be found on RTAB-Map's page at IntRoLab. Here a sample (better to watch in HD):
Older news AuthorOverviewRTAB-Map is a memory management approach for real-time appearance-based loop closure detection for autonomous robot (using a simple camera or webcam). The loop closure detection approach is fully incremental, starting with an empty memory. The robot builds his own representation of the environment by linking new acquired images with previous ones (e.g., detecting loop closures). The method implemented here can be considered as a Topological SLAM (Simultaneous Localization And Mapping) approach. Memory management makes it possible to process each new image under a fixed real-time limit, thus ideal for long-term operation and systems with limited resources.
More videos here. Under the hood : - Loop closure hypotheses are evaluated using a Bayes filter.
- Image matching is done using a "bag-of-words" approach (with SURF features - OpenCV).
- Memory is divided into short-term, working and long-term memories.
- Sqlite3 database is used for the long-term memory.
Try it!- For ROS users, take a look to RTABMap_ROS introduction and package overview.
Older newsMarch 2012 - Release 0.3.2:
- Simplified RTAB-Map's main parameters by adding a basic panel in Preferences.
- Added a new signature type (Sensorimotor) based on indexed colors, pixel to pixel comparisons for fast and precise sensorimotor learning (mostly used by the ros-pkg for learning).
November 2011 - I've made a simple interface to play with feature detectors and descriptors of OpenCV. See it at Find-Object.
July 2011 - New videos with different data sets are added to the home page.
June 2011 - The new version (0.3) of RTAB-Map is able to learn directly sensorimotor skills (only with the ROS packages).
- ROS packages are now integrated in this project (see RTABMap_ROS page).
- Tutorials!
|