First of all I am curious to see what everyone else is thinking about shooting accurately from a programmatic standpoint.
My plan was to use the axis camera to find the location of the retro-reflective tape rectangles using IR leds as a light source. Then I was going to center the rectangle on the camera frame and then measure the size and distortion of the rectangle to find location, and then use that to choose a correct trajectory.
We are heavily interested in the idea of successfully making shots based on the math of the entire thing, I’m interested in seeing how this will all develop in the coming weeks and will be sure to keep in touch, also I’ll be sure to check constantly to to see what everyone on CD is planning… Good Luck this season!!!
We are planning to find the target using a machine-learning based approach of classifying objects as noise, the target, robots, and other things, and using this data, we can acquire and track the target. Once we have the location of the target, we are using a regression to figure out where we are. We are not planning to illuminate the target for the sake of being able to use this anywhere on the field.
I would recommend taking a minute to play with illuminating the target. the retroreflective tape will shine like a star in the dark in terms of vision processing.
Why use a more complex solution when there is a simple one on the table for free?