基于改进支持向量机的视频摘要技术

Video Summarization Technology Based on Improved Support Vector Machine

鉴于视频监控系统获取到的视频数据十分庞大且其中存在大量冗余信息,通过视频摘要技术将这些视频数据浓缩成简短的摘要视频,让用户快速、准确地获得视频中的重要信息,提高视频数据的管理效率。采用二进制化的支持向量机和滑动窗口结合的方法,利用图像在HSV颜色域中的梯度信息,提取原始视频中具有重要信息的关键帧,然后把提取到的关键帧融合得到摘要视频。实验结果表明,该方法不但能够快速、准确地提取关键帧并形成摘要视频,而且能够取得较好的效果。

Aimed at the problem that video surveillance system obtained a huge amount of video data which exit a large number of redundant information,the paper proposed to condense these video data into a brief summary video by the video synopsis technology. The summary video can allow the user to get the important information in video quickly and accurately,so that the management efficiency of video data is improved. Using the means of combining the binarized support vector machine and the sliding window,we use the gradient information in the image in HSV color field,and extract the key frames which contain the important information in original video,then the summary video is gotten by fusing these key frames. Experimental results show that the method can not only extract the key frames quickly and accuracy from original video and get a summary video,but can also get good effect.