一种基于视频的篮球运动员能力评估方法

A Video-based Method for Evaluating the Ability of Basketball Players

为了更准确客观地对球员进行评估,文章提出一种基于篮球运动员的第一人称视频评价篮球运动员成绩的方法。通过利用第一人称摄像机获取球员的第一人称视频,并提供特定于评估者偏好的球员能力度量。该模型使用长短期记忆神经网络从第一人称视频中检测元篮球事件,并构造一个高度非线性的视觉时空篮球评估特征。实验结果表明,基于篮球竞赛标准即评估者的基本标准之一,提出模型学会了在真实的比赛中准确地评估球员,并且相对于现有的方法,提出的方法具有较好的性能。

In order to evaluate players more accurately and objectively,this article proposes a method to evaluate basketball players’performance from the first-person video of basketball players.The first-person video of the player is obtained by using the first-person camera,and provides a measure of player ability specific to the evaluator’s preference.This model uses a long and short-term memory neural network to detect meta-basketball events from first-person videos,and constructs a highly nonlinear visual spatiotemporal basketball evaluation feature.The experimental results show that based on the basketball competition standard,which is one of the basic standards of evaluators,the proposed model learns can accurately evaluate players in real games,and the proposed method has better performance than existing methods.