目标跟踪是计算机视觉中构建复杂运动分析系统的关键任务之一。体育运动比赛因其趣味性深受广大观众喜爱，对体育视频中的运动目标进行跟踪在赛后复盘、技术提升、教育教学以及媒体传播上都有着广阔的应用前景。近年基于深度学习的体育视频跟踪技术取得了显著进展。本文首先介绍了体育视频中目标跟踪研究背景，给出了基于深度学习的体育目标跟踪的定义与分类。其次分别从球类目标跟踪、单摄像机运动员跟踪以及多摄像机多运动员跟踪三个方面总结了体育视频中目标跟踪的研究现状，提炼出统一的算法流程并对其进行技术分析和总结。最后讨论了体育视频中视觉目标跟踪存在的挑战和未来发展方向。

The Object tracking is one of the key tasks in building complex motion analysis systems in computer vision. Sports competitions are deeply loved by audiences because of their fun. Tracking sports objects in sports videos has broad application prospects in post-match review, technology improvement, sports education and media communication. Significant progress has been made in sports video tracking based on deep learning in recent years. This paper first introduced the research background of object tracking in sports videos, and gave the definition and classification of objects tracking based on deep learning. Secondly, we summarized the research status of object tracking in sports video from three aspects: ball tracking, single-camera player tracking and multi-camera multi-player tracking, extract a unified algorithm process for every task, and conduct technical analysis and summary. Finally, we assessed the existing challenges and forecast the future directions of visual object tracking in sports videos.