Multi Object Tracking System form Video Streaming using Yolo

DeepSort, a cutting-edge multi-target tracking algorithm created exclusively for video surveillance datasets, is presented in this article. The system solves the drawbacks of traditional tracking techniques, including their lack of precision, fragility, and real-time tracking restrictions. DeepSort considerably increases tracking robustness under challenging environments, such as occlusion and fluctuating illumination conditions, by utilizing the benefits of YOLO (You Only Look Once) in high-precision target recognition and lowering dependency on detection results. The effective computation approach of DeepSort is one of its primary characteristics. To maximize computation performance, it includes a difference filter that filters out frames with no foreground targets or little changes. This method improves overall tracking speed while simultaneously lowering detection costs. DeepSort demonstrated outstanding performance in experimental assessments carried out on a sizable sports game video surveillance dataset. It had a maximum rate of 60 frames per second. Additionally, it obtained a highest Multiple Object Tracking Accuracy (MOTA) of 78.1% and MOTP of 80.9%. As in Sports analysis, player performance evaluation, and referee decision-making are all significantly impacted by the viability and efficacy of DeepSort in multi-target tracking. The algorithm’s precision and real-time capabilities can significantly improve the sports spectator experience by delivering accurate and continuous player and object monitoring during games. In conclusion, DeepSort presents a powerful multi-target tracking method for datasets from sports game video surveillance. DeepSort has the potential to revolutionize several facets of sports, including fan interaction, coaching, and officiating, because of its enhanced accuracy, robustness against occlusion, and effective processing method.