Embedded motion tracker using computer vision. This project envisages generic efficient motion tracking based on computer vision algorithms for resource limited devices.

Most dependencies are because of the diverse availability of capturing devices and video en- and decoders available. May be improved later.

For the image processing part:

• wget & unzip for retrieving third party dependencies.
• gcc or clang
• scons
• cmake
• ffmpeg + swscale (minimal dependencies of ffmpeg: avformat, avcodec, avutil, swscale)
• For Linux:
  • V4L (V1 & V2) (only for webcams that support V4L. Ximea and Xine webcam support can be enabled in the openCV-sconsbuilder settings).
• For MAC OS X:
  • Cocoa framework (normally standard present) => Ffmpeg can be replaced by gstreamer package (minimal dependencies: gstreamer, gstapp, gstpbutils, gstriff, gobject, glib2.0)

For GUI:

• python 3 (3.4 recommend, else change in demo/demo_macosx)
• scipy
• numpy
• matplotlib
• For Linux:
  • GTK2 (minimal dependencies: gtk-x11, gdk-x11, cairo, gdk_pixbuf-2.0, gobject-2.0, glib-2.0)
• For MAC OS X:
  • Qtkit (normally standard present)

All necessary packages are available in most Linux package managers and Macports.

Names of the necessary packages in macports:

• gcc 4.8 : gcc48 (it is recommended to install gcc using the 'Command Line Tools' (part of Xcode) from the App Store, this way you will get the latest clang compiler)
• python2.7 : python27
• python3.4 : python34
• ffmpeg : ffmpeg (gstreamer 1.0 : gestreamer1 and gst-ffmpeg)
• scipy: py34-scipy
• numpy: py34-numpy
• matplotlib: py34-matplotlib

run get3rdpartyGit.sh:

$ ./get3rdpartyGit.sh

run demo (first time the project will be built):

• for Linux:

  $ cd demo; ./demo_linux.sh

• for MAC OS X:

  $ cd demo; ./demo_macosx.sh

For localhost demo's run the demo as described in 'how to build'. Alternatively or for non-localhost usage, use the python wrappers in wrappers/. Use the -h function for more info on each wrapper.

NOTE: for the best results it is recommended to set the exposure time of the capture device to manual.

Contact bart.verhagen@tass.be