Efficient solution for simple scaled shape problem

asked 2013-03-19 22:16:12 -0600

I have images with 3 bright red (reflected laser) points defining the vertices of a constant triangle. The triangle will be significantly scaled and slightly translated in each image, but not at all rotated. All I want is the scale factor relative to some reference size.

I imagine this is a relatively easy problem in opencv but I'm hoping someone can point me to a very efficient algorithm. I need it to run on a low power CPU, maybe raspberry pi or similar, ideally processing 5Mp images at more than 10 fps, although I could settle for less resolution. Is this achievable? If so, algorithm suggestions would be very much appreciated.

edit retag flag offensive close merge delete

add a comment

1 answer

Sort by » oldest newest most voted

answered 2013-03-20 04:08:53 -0600

Guanta
6736 ●6 ●25 ●79

So basically you just need to detect these bright red spots. See the answer of http://answers.opencv.org/question/8818/point-tracking/ in which blue points are detected. Then, from any two points you compute the vector and consequently compare it with the vector of the same two points in your reference image for the scale factor.