基于视频图像阈值分割的网球运动员眼球运动轨迹捕捉优化研究

打开文本图片集

中图分类号：TP391 文献标识码：A

Optimization of Eye Movement Trajectory Tracking for Tennis Players Based on Video Image Threshold Segmentation

LI Tingwen1，2 ， ZHANG Jianhua

1.School of Physical Education，Minnan Normal University，Zhangzhou Fujian 3630oo，China；

2. Hubei Leisure Sports Development Research Center，Wuhan 430062，China；

3.School of Physical Education，Northwest Normal University，Lanzhou 730070，China

Abstract： Accurate eye movement trajectory tracking of tennis players remains challenging due to rapid eye motion and concentrated contribution of fixation points.The existing segmentation algorithms are dificult to accurately identify the trajectory due to local grayscale similarity，which affects the precision. To address this，we propose an optimized video image threshold segmentation method to enhance tracking accuracy and real-time performance for athlete training. Using GoPro HEROl2 Black cameras，we captured video data from professional tennis players at Minnan Normal University. The foreground images were obtained via Tsalis relative entropy-based multi-threshold segmentation and fed into a deep learning model combining convolutional neural networks and attntion mechanisms for feature extraction and optimization.A regularization term was introduced to mitigate overfiting and refine feature output. Experimental results demonstrate that our method effectively segmented the foreground images，producing clear and accurate eye movement trajectories that closely aligned with the ground truth，validating its superior threshold segmentation capability. The optimized model exhibits stable convergence，with loss values approaching zero with the training rounds，ensuring high efficiency and precision in trajectory tracking.

Key words： video image； threshold segmentation；tennis players；eye movement trajectory；tracking optimization

眼球运动主要分为平滑追踪（追随）、快速眼跳（扫视）和注视3种类型[1]。（剩余13222字）