基于逆光图像校正的竞技体操连接加分动作分割方法

打开文本图片集

文章编号：1003-6180（2026）02-0057-07

[中图分类号]TP391 [文献标志码]A

A Segmentation Method for Combined Bonus Actions in Competitive Gymnastics based on Backlit Image Correction

ZHENG Zhongwen 1 ， XU Zhiyang 2

（1. School of Physical Education， Fuyang Normal University， Fuyang 236037， China; 2. College of Physical Education， Zunyi Normal University， Zunyi 564100， China）

Abstract： A competitive gymnastics connection bonus action segmentation method based on backlight image correction （Backlight Image Correction Method） is proposed to compensate for local overexposure and underexposure caused by backlight， thereby improving the accuracy of action segmentation. This method corrects the dark and bright areas of backlight images and utilizes a multi-scale fusion pyramid to fuse these bright and dark regions. It employs Gaussian Mixture Model （GMM） background modeling to process the corrected images， obtaining a background model， and combines it with an improved ViBe algorithm for action segmentation. Experimental results demonstrate that this method can enhance the image information entropy and average gradient， effectively segment connection bonus actions， and achieve high boundary smoothness in action segmentation under different light intensities， providing theoretical support for the automation and scientific assessment of competitive gymnastics scoring.

Keywords： backlight image correction ; competitive gymnastics; bonus action ; segmentation method ; secondary illumination; ViBe Algorithm

竞技体操是一种综合性运动，要求运动员动作精度高、艺术表现力强。（剩余7041字）