Vision Processing: Automating Artifact Recognition

In DECODE, picking up the wrong color (Opponent’s Artifact) is a penalty or a wasted cycle. Your intake moves too fast for human reaction. You need Computer Vision.

Option 1: Color Blobs (OpenCV)

• How it works: “Is there a big blob of GREEN pixels in this rectangle?”
• Pros: Extremely fast (60FPS). Simple code (Core.inRange, Imgproc.findContours).
• Cons: Lighting changes can break it. A green wall might look like a green artifact.

Option 2: Machine Learning (TensorFlow)

• How it works: You train a neural network on 1000 images of Artifacts. It learns the shape and texture.
• Pros: Robust. Works in weird lighting.
• Cons: Slower. Requires training data.

Implementation Application

1. Intake Reject: If the camera sees the wrong color inside the intake, the motor automatically reverses.
2. Auto Select: The robot drives to the stack and automatically targets only the Purple one.

Conclusion

Start with OpenCV logic. It handles 95% of cases. Only upgrade to TensorFlow if lighting varies wildly.