Alejo's Tenuki

Actually, I made a similar program a couple of years ago. But also I never published it.

It’s hard to make it robust and error-proof while limitting the configuration procedure.

The camera should be positioned as high as possible, in best situation it should be above the board. If you take a low viewing angle stones can be in front of eachother. In other words, for the stones on the farest end of the board you should not check the middle of the intersection, but you should check a little bit higher to avoid seeing the stone in front of it.
BUT when there’s no stone at the intersection, that means you could spot the stone behind it instead. So, the viewing angle is really one big issue.

The second issue is detecting the board borders. In this clip 4 stones are used for the borders assuming linear interpollation of the intersections. But actually that’s not enough because most webcams have cheap lenses and bend the images to a spheral shape. Linear interpollation is not precise enough here. So I don’t believe that 4 points is enough for most webcams. This also depends on the distance of the webcam.

Then there’s the problem of capturing a game while it’s in progress. There’s a lot of issues there. Moving hands accross the board, should not be confused with stones! I tried to reduce this by ignoring big changes to the board. But actually, … when you capture big quantities of stones, that’s also a big change.

What if somebody suddenly turns on the light? All colors of the stones and the board suddenly change.

What if the camera moves during the game, or just shakes a little bit, … configuration should be performed again.

It’s hard to make it error-proof.