摘要
体育视频属于一种十分关键的信息资源,高精度分类体育视频可提高用户浏览与查询效率。针对当前分类体育视频结果主观性强、区分正确率低等缺陷,提出了基于深度学习的体育视频分类方法,采用相似系数关键帧提取算法获取关键帧特征,通过深度学习编码模型建立体育视频图像分类方法,以3类体育视频为例测试文中方法的性能,结果表明,对于不同类型的体育视频,文中方法的分类整体效果要明显优于当前其他体育视频分类方法,大幅度提升了体育视频分类效果。
Sports video is a very important information resource.High precision classification of sports video can improve the efficiency of browsing and querying.In view of the shortcomings of current sports video classification results,such as strong subjectivity and accuracy,a sports video classification method based on deep learning is proposed.The key frame feature is obtained by using the similar coefficient key frame extraction algorithm,and then the sports video image classification method is established by using the deep learning coding model.The performance of this method is tested by taking three kinds of sports video as examples.The results show that for all kinds of sports video,the classification effect of this method is obviously better than other sports video classification methods,which greatly improves the classification effect of sports video.
作者
肖友定
XIAO Youding(Physical Education Department of Shanghai Jianqiao University,Shanghai 201306,China)
出处
《电子设计工程》
2021年第3期162-166,170,共6页
Electronic Design Engineering
关键词
深度学习
体育视频
分类
关键帧
编码模型
微调
deep learning
sports video
classification
key frames
coding model
fine-tuning