It is divided into 3 parts: recognition system, control system, and mechanical part.
The recognition system include a webcam, and the program written with OpenCV.
The Control system include PC and Arduino.
The mechanical part is the actuator arm of the hard disk drive and the touch pen.
The webcam has low fps and high delay, so we use open-loop model to calculate the bird position. But the accumulative error of open-loop model causes instability. We will improve it in the future. The pipe in game is green, then we recognise the pipe with color filter, and correct camera distortion. We write record and replay tools with data visualization in Python to help us adjust the model.