0

I'm doing a project on hand posture recognition with OpenCV.

I've done the segmentation using normalizedRGB and found the contours using cvFindContours. Now I need to find the features to be extracted.

What are the best features to extract? What is the best method of classification for this case?

Thanks for the help ...

eboix
  • 5,113
  • 1
  • 27
  • 38
user1691622
  • 3
  • 1

1 Answers1

0

  1. if your hand is static...you can fit a polygon to your contour and try to charecterise the polygon according to your gesture...like the number of sides of the polygon...angle between the sides...area of A,B,C,D...as shown in following diagram... enter image description here

  2. if your hand is moving then you can use cvCalcOpticalFlow to track the hand and depending on the displacement between two consecutive frames charecterise the direction of movement and take some decision...

rotating_image
  • 3,046
  • 4
  • 28
  • 46