上海交通大学学报

上海交通大学学报（自然版） ›› 2014, Vol. 48 ›› Issue (05): 668-673.

• 自动化技术、计算机技术 • 上一篇    下一篇

基于两级支持向量机分类的视频镜头分割方法

俞璐1,2，乔瑞萍2，胡宇平2，赵静1,2
  

  1. （1. 宇航动力学国家重点实验室， 西安 710043; 2. 西安交通大学 电子与信息工程学院， 西安 710049）
     
     
     
  • 收稿日期:2013-06-24
  • 基金资助:

    国家自然科学基金青年科学基金项目（61102082），宇航动力学国家重点实验室开放基金项目（2013ADLDW0302）

A Video Shot Segmentation Method Based on a Two-Stage Support Vector Machine Classifier

YU Lu1,2,QIAO Ruiping2,HU Yuping2,ZHAO Jing1,2
  

  1. (1. State Key Laboratory of Astronautic Dynamics, Xi’an 710043, China; 2. School of Electronic and Information Engineering, Xi’an Jiaotong University, Xi’an 710049, China)
     
  • Received:2013-06-24

摘要：

提出一种基于两级支持向量机分类的视频镜头分割方法.第1级分类器利用分段视频首尾帧直方图距离，结合滑动窗口和陷波方法计算分段视频的特征向量，通过支持向量机来分类筛选含有镜头边界的子段；第2级分类器根据不同间距的帧间直方图的距离特征，采用时间窗口法构造特征向量，利用二叉树支持向量机多分类策略检测镜头边界的位置.结果表明，所提出的方法能够同时提高切变和渐变的镜头边界的检测效果.
 
 

关键词: 镜头分割, 两级分类, 支持向量机, 分段筛选

Abstract:

A video shot segmentation method based on a two-stage support vector machine (SVM) classifier was proposed. In the first stage，the feature vectors were generated by using the segmented video histogram distance, combining the sliding window with the trap method. Then, the subsegments containing shot boundaries were filtered out by using the SVM classifier. In the second stage，the feature vectors were obtained by histogram distance between different spacing of the frames and time window. The shot detection was implemented by binary tree SVM classification strategy. Experimental results show that the detection results can be improved on both abrupt and gradual shot boundary significantly by the proposed method.
 

Key words: shot segmentation, two-stage classifier, support vector machine (SVM), fragment screening

中图分类号: