1

I am working on a Kinect project that needs to recognizes flat surface and small objects. I am planning to use openni on a mac machine. So what libraries do I need to used for object recognition? What knowledges do I have to know to finish this project.

xinghua
  • 139
  • 2
  • 13

2 Answers2

2

Probably OpenCV would be a good start.

M Rajoy
  • 4,028
  • 14
  • 54
  • 111
2

If you're planning to do this in 3D (i.e. using the point cloud from the Kinect), you may find PCL useful. This C++ library has lots of functions for interpreting 3D point cloud data, such as RANSAC plane finding, normal computation, segmentation, feature extraction etc. See the tutorials page here for some examples of what can be done.

However, PCL does not contain any machine learning algorithms as yet, so you may wish to make use of some of those in OpenCV.

If you are more familiar with MATLAB, you may find it easier to do your work there, as you can process your data one step at a time and easily visualise your results as you go. However, there are fewer built-in 3D processing functions, and it will run much slower than a C++ program.

Bill Cheatham
  • 11,396
  • 17
  • 69
  • 104