Right now our team is trying to use grip fro image recognition (the reflective tape on the hoop) We’ve gotten the robot to recognize the tape, but its unable to recognize it from a distance greater than a couple of feet due to interference around the room. We are wondering if anyone has found algorithms or settings that work well?
What does your pipeline look like? Typically you look for only a certain color (like green)
GRIP is a great tool for image recognition in FRC. Usually, you want to find contours of a range of colors in an image. The steps my team(Steampunk 1577) uses are:
HSVthreshold → findContours → filterContours
If you can’t recognize your target from a large, reasonable distance it can be because of various reasons:
- your HSV threshold isn’t wide enough, try to optimize it to include also the color reflected from a larger distance.
- your LEDs aren’t bright enough.
- your camera’s settings aren’t for you. usually, when you want to recognize a reflective tape, you will set your camera’s exposure on a low value so only direct light will be counted. try to see If when looking at your regular stream, you can easily recognize your target.