摘要
为了提高运动视频图像分类精度,提出了利用多特征融合和最小二乘支持向量机的运动视频图像分类方法。采集运动视频,采用两帧差分法分割运动视频图像,并提取运动视频图像的多个特征;将不同特征输入到最小二乘支持向量机中进行学习和训练,获得单一特征的运动视频图像分类结果;利用证据理论对单一特征的运动视频图像分类结果进行融合,得到运动视频图像的最终分类结果。试验结果表明,该方法的运动视频图像分类精度高,分类时间明显少于经典方法,且抗噪声干扰能力得到大幅度提升,分类结果可以为运动视频管理提供技术支撑。
In order to improve classification accuracy of motion video images,a motion video image classification method based on multi-feature fusion and least squares support vector machine is proposed.Firstly,the motion video is collected,the two-frame difference method is used to segment the motion video image,and multiple features of the motion video image are extracted.Secondly,different features are input into least squares support vector machine to learn and train,and classification results of single feature motion video image are obtained.Finally,the classification results of motion video images with single feature are fused by evidence theory to obtain the final classification results of motion video images.The experimental results show that motion video image classification accuracy of the proposed method is high.The classification time is significantly shorter than that of the classical methods,and anti-noise ability is greatly improved.The classification results can provide technical support for motion video management.
作者
孔祥魁
樊翠红
Kong Xiangkui;Fan Cuihong(College of Education and Sports Sciences,Yangtze University,Jingzhou 434023,China)
出处
《南京理工大学学报》
CAS
CSCD
北大核心
2022年第2期164-169,176,共7页
Journal of Nanjing University of Science and Technology
基金
湖北省教育厅哲学社会科学研究项目(20Y039)
长江大学教师教育研究中心创新基金资助项目(2020JSJYZX04)。
关键词
多特征融合
最小二乘支持向量机
运动视频
图像分类
证据理论
multi-feature fusion
least squares support vector machine
motion video
image classification
evidence theory
