Low Goal Tracking

[btcshields6]

Anybody thinking of tracking the low goal? Thoughts on how you might do it? I'm thinking maybe using the shape, since there's no vision tape mounted down there.

Just for ease of shot alignment.

[rich2202]

I am going to post a q&a asking for low goal vision target

[btcshields6]

Okay. I'd be surprised for them to respond positively and actually change part of the field per request, but that'd be great!

[RufflesRidge]

Originally Posted by btcshields6

Thoughts on how you might do it?

Track the high goal that is directly above it?

[Greg McKaskle]

You may also be able to recognize the texture of the tower and interruption in that texture. I think many texture training tools are size dependent, but corner detectors, especially at a coarse pyramid level may be a reasonable approach.

The carpet color should be pretty predictable, and the tower ramp color may be as well, depending on lighting. You also have lighting strips on the corners of the tower which will often be there in your opponent's alliance color.

Greg McKaskle

[Fauge7]

Either go with greg, or simple go for the vision tape up top

[btcshields6]

Thank you all for your suggestions, especially the in-depth ones.

I guess we'll see as the season goes on.

[JCharlton]

While I can't say I've tried it with the actual field, this is a solution you might try. For my example I used a screen capture from the field tour video.

After you acquire the image the algorithm works like this:
- Crop the image You probably will capture more than you want. Get rid of as much of the background as you can.
- Search for dark parts of the image with colour threshold Use HSL and filter with luminance. You don't need the full colour threshold function, but it works for other things too.
- Filter out small bits with Particle Filter 3 This should clean up a lot of junk.
- Dilate the image to merge particles into one The intention of this step is to blur all the particle blobs into one big particle.
- Use particle analysis You'll want to a) check there's only one particle, and b) get the x position.

Not saying it'll be easy, but it might give you something to experiment with, though the suggestions to look for the upper vision target are valid too.