1. Download source (153KB) (prepared for visual studio 2010. For other versions, please prepare the solution file)
  2. Set up OpenCV (2.0 or later) and OpenGL GLUT
  3. Add OpenCV headers and library paths in mylib/opencvpath.h
  4. Set up OpenMP (optional)
  5. Compile the source
  1. Print markers in data directory (like 10cm x 10cm)
  2. Run the software
  3. Click 's' to watch binary images and check black dots are extracted or not
  4. Click ESC to close the software and change the threshold for binaization at Binarization() in mylib/myimage.cpp
  1. Run the software
  2. Show markers to a camera (rectangles are displayed if markers are successfully retrieved)
  3. Click ESC to close the software (takes time for releasing memory)
  1. Download one thousand markers (13MB) and put them into data directory (download markers in PDF for printing if need)
  2. Make your own markers with our code
  3. Set the number of markers you use on int nummarker at trackingInit() in main.cpp (need 500MB for execution if use 1000 markers)
  1. Keypoint matching by locally likely arrangement hashing (LLAH)
  2. Multiple planar object retrieval and tracking
  1. Better to move the data directory into the directory of executable and directly run with the executable (running on visual studio is a little slow)
  2. Better to change the number of iterations for homography estimation with RANSAC in OpenCV for fast retrieval (need to change const int maxIters like 300 in cvFindHomography and recompile OpenCV)
  3. Use this software for only research (non-profit) purposes
  4. Feel free to email me for any questions
  5. Try deformable version
Back to top

Updated on Nov. 4, 2011.