All geek questions in one place

Shape recognition after edge detection

I am working on my first project on computer vision, where I draw a diagram and find symbols on it. There are 5 different forms that need to be discovered.

This is the result of my border detection algorithm: http://i.imgur.com/mh9Ol.jpg . 5 characters is one point, 4 points grouped together, a rectangle and 2 odd-shaped characters.

At this moment, Iā€™m not sure which algorithms I should look at, which will allow me to select these figures and distinguish them from each other. Any ideas would be very helpful. Thanks!

+6
source share
2 answers

It seems that Shape Context would be a natural choice for this type of problem.

+1
source

a few years ago I tried pattern matching with OpenCV, which is described here.

This worked very well, and I think this is what you are looking for; however, I remember that it was rather slow, so it may not be very convenient if you are programming a real-time application. If so, you may have to play around with the geometry of your shapes and try to find them using Hough transforms for lines and circles , both of which are implemented in OpenCV.

Good luck

0
source

Source: https://habr.com/ru/post/918901/

More articles:


All Articles