基于SVM的视频检索系统框架

A Video Retrieval System Framework Based on SVM

底层特征到高级语义的转化是视频检索中的重要环节。结合在视频检索中的相关反馈,结果排序以及BOOST思想,提出了针对视频检索的改进的SVM算法流程,并根据改进的SVM算法提出了基于SVM的视频检索系统框架。通过采用TREC2007的数据测试表明,该视频检索系统框架可达到较高的查准率和查全率。

Transformation from feature to semantic info is the important step of Video Retrieval. Based on the improved SVM procedure for Video Retrieval according to relevance feedback, reranking and BOOST, this paper proposes a framework. Using the data of TREC2007, the experimental result shows the framework has sufficient improvement in AP and MAP.