ProCamTracker

Introduction

ProCamTracker is a user-friendly computer application to turn a perfectly normal pair of color camera and projector into a system that can track without markers a real world object (currently limited to matte planes), while simultaneously projecting on its surface geometrically corrected video images using the direct image alignment algorithm included in JavaCV, an open source library I developed as part of my doctoral research. More information about the algorithm itself can be found in the related CVPR 2010 paper below, which you may cite if you find this software useful. Thank you.

Samuel Audet, Masatoshi Okutomi, and Masayuki Tanaka. Direct Image Alignment of Projector-Camera Systems with Planar Surfaces. The 23rd IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2010). IEEE Computer Society, June 2010.

This method requires a geometrically and color calibrated projector-camera system. To perform the calibration, I recommend ProCamCalib, a tool I previously released for that purpose.

More information at http://www.ok.sc.e.titech.ac.jp/res/PCS/research/procamtracker/

Downloads

Archives containing JAR files are available as releases. The binary archive contains builds for Linux, Mac OS X, and Windows.

Required Software

I wrote ProCamTracker itself in Java and its binary should run on any platform where an implementation of Java SE 7 or newer exists. The binary distribution also contains natively compiled code for Linux, Mac OS X, and Windows, needed by JavaCV.

Please install the following before running ProCamTracker:

As well as the following to enable processing with OpenCL and OpenGL:

And be aware that ProCamTracker runs a lot faster under the "server" JVM than the "client" JVM, but because of its bigger size, not all distributions of Java come with the server one.

Additionally, for IIDC/DCAM cameras, Microsoft's Kinect stereo camera, or the PS3 Eye:

Usage

Under Linux, Mac OS X, and other Unix variants, execute either procamtracker-nativelook or procamtracker-oceanlook, according to the theme that works best on your system. ("Ocean" being Java's original look and feel.) The equivalent files under Windows are procamtracker-nativelook.cmd and procamtracker-oceanlook.cmd.

After launch, the user interface that appears allows the user to change settings for the camera, the projector, and the various modules of the tracking algorithm. I describe next a typical usage scenario and will not explain all the settings in detail. The default values of the ones I do not mention should be good enough for most cases, but please feel free to experiment.

  1. Camera Settings
    Select a suitable FrameGrabber for your system, and fill in either deviceFile, deviceNumber or devicePath, and format as appropriate. Place the path to the calibration.yaml file created by ProCamCalib in the parametersFile field, while the name field must correspond to an entry in that file.
  2. Projector Settings
    As with the camera settings, fill in the parametersFile field, but also confirm that the screenNumber corresponds to the one of the projector.
  3. RealityAugmentor/ObjectSettings/VirtualSettings
    Locate an image file (PNG, JPG, etc.) or video file (AVI, MP4, etc.) you would like to project on some planar surface and specify its path in the projectorImageFile or projectorVideoFile field respectively.

Once you have modified all the desired settings, since the application may crash during the operations described below, please save them in an XML file via the Settings menu.

Now, we are ready to start the tracking algorithm. From the Tracking menu, click on the Start item. If all goes well, a window entitled "Initial Alignment" should appear containing an image captured from your camera. In this window, click four different points that delimit the region of interest on your planar surface that you would like to map to the four corners of the projector image, in this order: upper left, upper right, lower right, and lower left. After the last click, tracking should start within a few seconds. You may then begin to move your planar surface and see if the system can successfully track it.

Feel free to contact me if you have any questions or find any problems with the software! I am sure it is far from perfect...

Source Code

I make all the source code available on GitHub at https://github.com/bytedeco/procamtracker . You will also need the following to modify and build the application:

(The icons were shamelessly copied from the source code repository of NetBeans. Also licensed under the GPLv2.)

Please keep me informed of any updates or fixes you make to the code so that I may integrate them into my own version. Thank you!

Project lead: Samuel Audet samuel.audet at gmail.com
Developer site: https://github.com/bytedeco/procamtracker
Discussion group: http://groups.google.com/group/javacv