Efficient Ball Position Estimation for Tennis Court Robot Assistants using DualCamera System

Abstract

During tennis training, professional players use many balls that will be distributed randomly around the tennis court.Collecting the ball can be done manually, However, the effort and the time make it inefficient.The mobile robot for ball collection has been introduced as a solution to save power and training session time.The robot tasks involve several steps: ball detection, estimating their positions, and finding the best path for efficient collection.In our previous work, we addressed the ball detection issue utilizing a neural network algorithm using YOLOv8.This study focuses on the next step, ball position estimation, using two cameras to cover the entire court.The results demonstrate successful ball position estimation along the x-axis and y-axis, achieving an accuracy of 97.76 %.