新媒体背景下的视频广告分类系统设计  

Design of Video Advertisement Classification System under the Background of New Media

作　　者：吴韵怡 WU Yunyi(School of Arts and Media,Guangzhou Huali Science and Technology Vocational College,Guangzhou 511325,China)

机构地区：[1]广州华立科技职业学院,艺术与传媒学院,广东广州511325

出　　处：《微型电脑应用》2022年第4期65-68,共4页Microcomputer Applications

基　　金：广东省教育厅普通高校青年创新人才类项目(2021WQNCX271)。

摘　　要：为了获得理想的海量视频广告检索结果,提出新媒体背景下的视频广告分类系统。该系统依据基础层反馈的检索需求,逻辑分析与处理层对原始视频广告序列进行镜头分割,采用多支流网络对中间帧序列实施编码,提取视频广告中的有用信息,向多支流网络的各支流引入空域注意力预估模型,融合注意力谱和各条支流网络提取的特征谱,通过两个全连接层获取视频广告的分类评分,并线性叠加,获取整体视频广告的类型预估结果,将该结果反馈给用户。结果表明,该系统可实现视频广告分类,分类精度高,系统整体性能优越。In order to obtain ideal mass video advertisement retrieval results,a video advertisement classification system under the background of new media is proposed.According to the feedback of the basic layer,the logic analysis and processing layer segments the original video advertisement sequence,uses multi tributary network to encode the intermediate frame sequence,extracts the useful information in the video advertisement,introduces the spatial attention prediction model to each tributary of the multi tributary network,integrates the attention spectrum and the feature spectrum extracted from each tributary network.The total connection layer obtains the classification score of video advertisement,and linearly superimposes it to obtain the type estimation result of the overall video advertisement,and feeds back the result to the user.The results show that the system can achieve video advertising classification,classification accuracy is high,and the overall performance of the system is superior.

关 键 词：新媒体 视频广告 镜头分割 视频帧 特征提取 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]