地球信息科学学报 ›› 2012, Vol. 14 ›› Issue (6): 686-692,697.doi: 10.3724/SP.J.1047.2012.00686

• 地球观测数据获取与处理 • 上一篇    下一篇

区域人群状态的实时感知监控

宋宏权, 刘学军, 闾国年, 张兴国   

  1. 南京师范大学 虚拟地理环境教育部重点实验室, 南京 210023
  • 收稿日期:2012-11-01 修回日期:2012-12-10 出版日期:2012-12-25 发布日期:2012-12-25
  • 作者简介:宋宏权(1986-),男,河南民权人,博士研究生。研究方向:视频GIS,地理信息系统设计、开发与应用。E-mail:hongquansong@126.com
  • 基金资助:

    国家"十二五"科技支撑计划课题"视频GIS与突发公共事件的感知控制系统"(2012BAH35B02);江苏省高校自然科学重大基础研究资助项目(10KJA420025)。

Real-time Monitoring for the Regional Crowds Status

SONG Hongquan, LIU Xuejun, Lü Guonian, ZHANG Xingguo   

  1. Key Laboratory of Virtual Geographical Environment, Ministry of Education, Nanjing Normal University, Nanjing 210023, China
  • Received:2012-11-01 Revised:2012-12-10 Online:2012-12-25 Published:2012-12-25

摘要:

公众聚集场所人群高度聚集,流动性大,隐藏着巨大的安全隐患,时常发生群死群伤的拥挤踩踏等突发公共事件。针对现有以视频监控的人群分析,无法从空间视角掌握区域人群状态的时空格局,本文提出了面向人群分析的视频GIS框架,将视频数据映射至地理空间,在GIS环境下提取人群密度、人群运动矢量场等人群特征。通过分析人群运动矢量场可得到人群运动模式及各方向人群主体运动速率。最后,将视频监控系统与GIS进行有机集成,设计并实现了以视频与GIS协同的区域人群状态实时感知监控系统。实验结果表明,本系统可为大型集会活动的突发事件预防、人群疏导等提供决策依据。

关键词: 感知, 视频GIS, 监控, 时空格局, 人群监测

Abstract:

With the rapid development of the social economy, the massive crowd gathering appears frequently. Personnel casualties often caused by higher crowd density. So, video surveillance technology has become a national policy in many countries. Surveillance cameras have been installed in various important places of the city. Real-time monitoring of the crowds status in crowd gathering area can provide important basis for crowd management and emergency warning. Existing video-based crowd analysis can only monitor crowd status for each camera separately. We cannot get the spatial-temporal patterns of regional crowd status from a spatial perspective. In this paper, we proposed a video-GIS framework for crowd analysis. Video frames can be mapped to geographic space based on the video-GIS framework. So we can process crowd images and extract crowd density, crowd movement vector field in GIS. Then the crowd movement pattern and the main direction of crowd movement can be acquired by the vector field analysis. Finally, we design and implement a real-time monitoring system for the regional crowd status using video surveillance system and GIS. Experimental results show that: (1) previous crowd analysis methods based on the image space can only measure results by the unit of pixels. It requires further conversion if we want to get the real value. But we can get the real value directly when we process crowd images in GIS using the method we proposed. (2) The accuracy of the pixel-based low-density crowd counting estimation results can be up to 90%. The classification accuracy of the high-density crowd levels support vector machine classifier is more than 95%. So, they can fully meet the needs of crowd monitoring. (3) We can get the crowd movement pattern and the main movement direction by the analysis of crowd movement vector field in GIS. Also, we can obtain the speed of the crowd in different directions. These crowd characters all can be expressed in GIS. (4) The system we developed for the crowd monitoring can be applied to crowd management and emergency warning. It can provide decision making basis for emergencies prevention and crowd divert.

Key words: crowd status monitoring, spatial-temporal pattern, real-time monitoring, perception, video-GIS