Arrow detection what is the best approach?

asked 2013-09-26 14:01:35 -0500

WG- gravatar image

updated 2013-09-26 14:27:03 -0500

I need to detect a arrow shape e.g.

image description

(the arrow will most likely have a different colour probably red)

This will be in a live environment i.e. the arrow will be printed on a piece of A4 paper and hung up in a corridor. The camera which needs to detect the image will most likely be a bit shaky so there will be some deformation of the image I presume, also lighting might be an issue. Further the arrow only needs to be detected from the front e.g. not from the side where it will be deformed.

I am now wondering now what my best approach might be to correctly detect the arrow and as such its direction, left or right.

Might for example this approach be a good start? http://stackoverflow.com/questions/15277323/opencv-shape-detection. Detect all the contours and then find all the heptagon's (7 vertices) in the result and then invent my own algorithm to check if one of the found heptagon represents a arrow? Or http://stackoverflow.com/questions/11424002/how-to-detect-simple-geometric-shapes-using-opencv using a approximates of polygonal curves?

Or might there be a approach/function where I give a image or shape of the arrow to the algorithm and then searches for similar objects/shapes in the image retrieved from the webcam.

edit retag flag offensive close merge delete

Comments

I can only suggest to consider also cascade classifiers.

GilLevi gravatar imageGilLevi ( 2013-09-27 06:48:12 -0500 )edit