摘要
特征选择是视频字幕定位的关键,为了提高视频字幕定位正确率,提出一种人工鱼群算法(AFSA)和最小二乘支持向量机(LSSVM)相融合的视频字幕定位模型(AFSA-LSSVM)。首先提取视频字幕特征,然后通过模拟鱼群的觅食、聚群及追尾行为找到最优视频字幕特征子集,最后将最优视频字幕特征子集输入LSSVM进行学习,建立最优视频字幕定位模型,并进行仿真对比实验。结果表明,相对其他视频字幕定位模型,AFSA-LSSVM提高了视频字幕定位正确率和效率,可为后续视频内容的安全分析提供技术支持。
Feature selection is a key problem in video caption location,in order to improve location rate of the video caption, a novel video caption loca-tion model （AFSA-LSSVM） is proposed in this paper which integrates artificial fish swarm algorithm （AFSA） with least squares support vector machine （LSSVM）. Firstly, video caption features are extracted, and then the optimal features subset are obtained by simulating feeding,clustering and the follow-ing behavior of fish warm,finally ,LSSVM is used to establish the optimal video caption location model based on the optimal features subset,and the simu- lation experiment is carried out to test the performance of model. The results show that compared with other video caption location models ,the proposed model has improved location rate and efficiency of video caption location,which can provide technical support for following video content security analy-sis.
出处
《电视技术》
北大核心
2014年第5期42-45,57,共5页
Video Engineering
关键词
字幕定位
特征提取
人工鱼群算法
最小二乘支持向量机
caption location
feature extraction
artificial fish swarm algorithm
east square support vector machine
