University Archives

During tennis training sessions, coaches typically hit dozens of balls toward players for practice. By the end of the session, players often need to stop training several minutes early to manually collect the scattered tennis balls, which reduces effective training time and interrupts the practice flow. To address this problem, this project proposes an autonomous robotic system designed to detect and collect tennis balls efficiently. The system uses an Intel RealSense D415 depth camera, which provides both RGB images and depth information to understand the 3D environment. A YOLO (You Only Look Once) object detection model is applied to identify tennis balls in real time and determine their locations within the scene. Using the depth data from the camera, the system estimates the spatial coordinates of each detected ball. These coordinates are then sent to the robot, which computes an efficient path and autonomously navigates the area to collect the balls. The proposed solution aims to automate the ball collection process, reduce interruptions during training sessions, and increase overall practice efficiency through the integration of computer vision, depth sensing, and robotic path planning.