OpenCV "Almost" Closed Loops

Question

I am trying to extract a cube from an image (looks like a square ...). I used canny and expanded to get the edges and remove the noise.

I'm not even sure if it is possible to get a square in a reliable manner.

Tips appreciated!

Thanks.

Answer 1

This is not too complicated.

Sort all edges in the direction. Look for a pair of ribs in one direction with the rotation of the other pair 90 degrees. Check the approximate proximity. If so, they probably form a rectangle. Check the distance between the edges to select squares from the rectangles and discard small squares. Check to see if you have large enough parts of the edge to make sure that the entire edge must exist. It is possible that the edge can be split into 2. Check if 4 edges are now found that have a fairly uniform area.

The last bit is a bit more complicated. This domain knows. Can there be other objects inside the square, and can they touch or overlap the edges of the square?

Answer 2

You can use kmeans color information and clustering as described in the link .

As long as the color of the target object is different from the background, the pixels of the square object can be detected accurately.