Abstract: Table tennis is a very rapid movement in competitive sports, and its referee needs to make a judgment based on observation and experience within a few seconds, which is a very complex task. Due to the subjective judgment of the referee’s uncertainty and imprecision, it will lead to some false judgment, resulting in a significant impact on the results of the competition. In the current competitions, the most controversial penalty is the height and angle of the throwing ball’s problem. To solve this controversial problem, this paper designs a complementary video-based recognition of table tennis service umpiring system that can identify and track the location of the table tennis in the video and determine the exact moment when the ball leaves the hand. As a new umpiring system, this paper first introduces into the machine learning methods to identify in table tennis, and discusses a new method of predicting the region of interest to reduce the recognition error rate and improve the recognition efficiency. This system will eventually get through the data of height and angle of the throwing ball, which not only provides a scientific reference for the referees, but also provides a supplementary basis for the athletes to question the referee.

Key words: table tennis, umpiring system, region of interest, machine learning, recognition

摘要： 乒乓球作为竞技运动中一类很快速的运动，其技术合法性需要裁判在几秒钟内根据观察和经验做出判断，这是一个很复杂的任务。由于裁判主观判断的不确定及不精确性，会导致比赛中产生一些误判现象，从而对比赛结果产生很大影响。在目前的比赛当中，争议最大的判罚即为运动员发球的抛球高度和抛球角度的合法性问题。为了解决这个争议问题，设计了一种辅助性的基于视频识别的乒乓球发球裁判系统，可以识别并跟踪视频中乒乓球的位置，并准确地确定乒乓球离开执拍手的瞬间。作为一种新型的裁判系统，第一次将机器学习的方法引入乒乓球识别中，并讨论了一种设置预测感兴趣区域的方法来降低识别错误率，提高识别效率。系统通过视频处理最终将得到运动员发球的抛球高度和抛球角度，力图在训练中为运动员提供一定技术参考，在比赛中为裁判员的判罚提供一定的科学参考，也为运动员质疑裁判的判罚提供一种辅助性的依据。

关键词: 乒乓球, 裁判系统, 感兴趣区域, 机器学习, 识别