欢迎访问《图学学报》 分享到：
图学学报

图学学报

• 计算机图形学与虚拟现实 • 上一篇    下一篇

一种基于螺旋摘要的监控视频可视分析系统

  

  1. (1. 中国石化销售股份有限公司华南分公司，广东 广州 510000； 
    2. 中国科学院大学计算机科学与技术学院，北京 100190；
    3. 中国科学院软件研究所人机交互北京市重点实验室，北京 100190)
  • 出版日期:2020-04-30 发布日期:2020-05-15
  • 基金资助:
    国家自然科学基金项目(61872346)

Surveillance video analysis system based on SpiralTape summarization

  1. (1. South China Branch of Sinopec Sales Co., Ltd., Guangzhou Guangdong 510000, China; 
    2. School of Computer Science and Technology, University of Chinese Academy of Sciences, Beijing 100190, China; 
    3. Beijing Key Laboratory of Human-Computer Interaction, Institute of Software, Chinese Academy of Sciences, Beijing 100190, China)
  • Online:2020-04-30 Published:2020-05-15

PDF (PC)

204

摘要： 监控视频是安防系统的重要组成部分。在如今的各行各业中，只要涉及到安全，均 离不开监控视频。但对监控视频内容的分析主要依靠大量人工来完成，人力和时间成本巨大。随 着监控视频数据越来越多，如何提高针对视频内容的分析效率、降低用户认知负荷是拓展视频利 用率的重要方面。为此，针对监控视频存在的冗余信息较多、人工获取视频关键内容效率低的问 题，采用螺旋视频摘要及相应交互技术，开发了一种面向监控视频内容的可视分析系统，结合运 动目标检测结果数据，基于螺旋摘要的展示优势实现多角度可视化视频目标统计信息，并辅以针 对螺旋摘要的导航、定位操作以及草图交互等方式，实现对监控视频内容的快速有效获取。

关键词: 监控视频, 螺旋摘要, 草图交互, 目标检测, 运动目标检测

Abstract: Surveillance video is an important part of security systems. In today’s all walks of life, surveillance video is inseparable when it comes to security. However, the video content is still analyzed largely manually, which costs much manpower and time. With the increasing amount of surveillance video data, it is important to figure out how to improve the analysis efficiency of surveillance video content and reduce user’s cognitive load in order to make higher use of the videos. Aiming at the problems of redundant information in surveillance video and inefficient manual acquisition of key video content, a visual analysis system for surveillance video content was developed by applying SpiralTape summarization and corresponding interactive technologies. Combined with the detection data of moving objects, the object statistical information of visual videos was obtained from multiple angles based on the display advantages of the SpiralTape summarization. Finally, fast and efficient access to surveillance video content was achieved with auxiliary methods such as navigation, positioning and sketch interaction for the SpiralTape summarization.

Key words: surveillance video, SpiralTape summarization, sketch interaction, object detection, moving object detection