摘要
为了提高体育视频识别的精度,提出一种多特征和神经网络相融合的体育视频识别模型(MF-RBFNN).分别提取反映体育视频的静态和动态特征,然后采用RBF神经网络对静态和动态特征分别分类,并将初步识别结果构造基本概率指派,运用证据理论对初步结果进行融合,得到体育视频识别结果.结果表明,相对于对比模型,MF-RBFNN提高了体育视频识别精度,是一种有效的体育视频识别方法.
In order to improve recognition rate of sports video,a novel sports video recognition method based on multi- features ad neural network was proposed. Firstly,the color,texture,brightness,motion vector features of sports video were extracted,and then the features were input into RBF neural network to learn and got the preliminary classification results which were taken as evidences of evidence theory,finally,evidence theory was used to fuse preliminary classification results and get the final recognition results of sports video,the simulation results show that the proposed method has improved the recognition rate of sports video and it is an effective sports video recognition method,compared with the reference algorithms.
出处
《湖南科技大学学报（自然科学版）》
CAS
北大核心
2015年第4期97-102,共6页
Journal of Hunan University of Science And Technology：Natural Science Edition
基金
湖北省教育厅项目(14G179)
关键词
体育视频
特征提取
证据理论
分类器设计
sports video
feature extraction
evidence theory
classifier design
