基于改进全卷积神经网络的体育运动员动作识别方法
Sports Athlete Action Recognition Method Based on Improved Fully Convolutional Neural Network
郝俊峰1
作者信息
- 1. 山西省中医学校,山西 太原 030012
- 折叠
摘要
传统的体育运动员动作识别方法,直接对运动员动作识别结果进行输出未对运动区域进行提取,识别精度低.该文提出基于改进全卷积神经网络的体育运动员动作识别方法,使用摄像机对体育运动员动作图像进行采集,并对图像进行基于改进全卷积神经网络的运动区域提取,体育运动员动作识别流程,输入动作图像并对结果进行输出,实现基于改进全卷积神经网络的体育运动员动作识别.实验结果表明该研究方法识别精度高,具有一定优势.
Abstract
Traditional sports athlete action recognition methods directly output the results of sports athlete action recognition without extracting the movement area,resulting in low recognition accuracy.The article proposes a sports athlete action recognition method based on an improved fully convolutional neural network.By using a camera to capture images of sports athlete movements and extracting motion regions based on an improved fully convolutional neural network,the process of sports athlete movement recognition is carried out.The input action image is then outputted to achieve sports athlete movement recognition based on an improved fully convolutional neural network.The experimental results indicate that the research method has high recognition accuracy and certain advantages.
关键词
改进全卷积神经网络/体育运动/动作识别/识别方法Key words
improving fully convolutional neural networks/sports activities/action recognition/recognition methods引用本文复制引用
出版年
2024
NETL
NSTL
万方数据