vision: chamfer_matching | cmvision | color_calib | cv_mech_turk | image_segmentation | image_synthesizer | people | place_recognition | pytoro | scanmatcher | toro | visodo | vslam_demo

Package Summary

Place recognition using a vocabulary tree.

• Author: Patrick Mihelich
• License: BSD
• Repository: https://code.ros.org/svn/wg-ros-pkg

GRAVEYARD

This package is currently in the graveyard until it is ported to work with the new visual odometry libraries. It and its dependencies can be checked out from:

https://code.ros.org/svn/wg-ros-pkg/graveyard/

Data files

The place_recognition and visual_odometry packages are research code; apologies if they are difficult to work with in their current form.

To use this package you will need a vocabulary tree trained on a large dataset of images. We have one here: http://vault.willowgarage.com/wgdata1/vol1/jbowman/holidays.tree

This tree was trained on the C-descriptors devised by Michael Calonder (calonder_descriptor). This descriptor has its own training stage which produces a random forest classifier. The classifier used to generate the above tree is available in two ways:

• Do roscd calonder_descriptor && make test. This will automatically download the classifier current.rtc to the calonder_descriptor package directory.

• Download it directly from http://pr.willowgarage.com/data/calonder_descriptor/current.rtc

Tutorials

Links to relevant tutorials