Actracktive: A highly configurable real-time computer vision platform

The last two years I have been working with interactive tabletops, like the one of my colleague Johannes Luderschmidt. One thing we more or less constantly struggle with is to get the touch/fiducial tracking working properly. As all our setups involve camera-based, optical tracking, this all comes down to doing image processing and applying computer vision algorithms. Of course, there are already several tools out there trying to address those issues. However, they did not work for us, so I started to develop our own software for this purpose: Actracktive.

Actracktive is a highly configurable real-time computer vision/image processing platform written in C++. It provides a framework and runtime environment for developing and executing complex processing graphs which can for example capture images from a camera, pre-process them using various image filters, extract features, track the detected objects of interest over time and finally send information about the objects to another application via network.

It originally started as a fork of the CCV project used for vision-based finger tracking in the multitouch/interactive tabletop systems. As a more modular and configurable solution was needed, a new architecture and implementation was built from scratch. It aims at providing a general and easily extensible framework for implementing most computer vision tasks (2D and 3D), but also other kinds of (frame-based) signal processing.

I now published the whole project on Google Code and made it open source (under the GPLv3 license). If you are interesed in this project, feel free to go over to the

Actracktive project page on Google Code

Leave a Reply

You must be logged in to post a comment.