Implementation and Analysis of a Computer Vision-Based Ball Detection System for Robot Soccer in ROS Simulation

This research investigates a computer vision-based ball detection system designed for robotic soccer applications, with a focus on assessing its efficiency across varying distances from 1 to 12 meters. The study utilizes the YOLOO (You Only Look Once) algorithm to evaluate the robot's detection accuracy, response times, and navigational performance in real time. A dataset composed of 3,400 images was created to facilitate the training and validation of the detection system under diverse lighting conditions. Findings revealed that the system achieved optimal detection accuracy of 70% at 1 meter, which decreased to 31% as the distance increased to 12 meters. The average response time for detection was recorded at 120 milliseconds for close distances, escalating to 228 milliseconds for farther distances. These results indicate the necessity for further enhancements in computer vision capabilities to improve interaction in dynamic settings such as robotic soccer. The study underscores both the potential and limitations of current detection methodologies, offering pathways for future advancements in robotic automation within competitive environments.