Deep Learning-based Curling Detection and Trajectory Tracking
Curling movement track capture can digitally restore the curling motion curve,which is not only conducive to the au-dience to better understand the competition situation,but also help players to make a judgment on the current situation.Robustly tracking a moving stone from curling video sequences is difficult because the stone is frequently hidden by the brushes held by the players and the players'bodies when players interact with stone.By optimizing the detection algorithm in deep learning,this paper realizes the capture of curling objects,the fast calibration of coordinates and the synthesis of curling trajectories.Multi-target track-ing method combined with curling motion characteristics and curling appearance.This paper also provides an actual curling dataset,which makes up for the lack of large-scale data set of curling.Compared with the existing methods,the experimental results show that the proposed method can efficiently perform tracking task with high accuracy and recall.
国家科技期刊平台
NSTL
万方数据
