Low-cost Ping Pong Ball Pickup Machine Based on Raspberry Pi

This paper proposes a vision-based ping-pong ball grasping system using Raspberry Pi, which utilizes a high-definition camera and a robotic arm to achieve precise grasping of ping-pong balls. The system aims to achieve low-cost ping-pong ball retrieval, reducing the need for manual picking. Firstly, the system achieves precise recognition of ping-pong balls by adjusting the HSV values of the balls, followed by extracting the contours of the ping-pong balls using Hough Transform. To convert pixel coordinates to robotic arm coordinates, the system fixes the camera height to transform three-dimensional coordinates into two-dimensional coordinates, reducing computational complexity. By employing camera intrinsic and extrinsic matrices and Zhang's calibration method, the system realizes the conversion between camera image positions and real positions. Through robotic arm hand-eye calibration, the transformation matrix for converting between real coordinates and robotic arm coordinates is obtained. Finally, noise reduction techniques such as Gaussian low-pass filtering are applied to improve system accuracy.