一种可跨摄像机的人群密度估计模型

A Cross-camera Adaptive Crowd Density Estimation Model

为实现人群密度估计模型的跨摄像机应用,提出一种模型构建方法。利用地理信息系统(GIS)技术,将视频数据统一至GIS空间坐标系,在GIS中实现人群图像的尺度统一,在GIS空间构建模型以实现其可跨摄像机应用,并将视频监控系统与GIS有机集成,研发人群密度实时感知系统。研究结果表明,基于提出的模型构建方法,利用某摄像机数据构建的模型能被应用于其他摄像机监控场景,无需为各摄像机构建模型。所研发人群密度实时感知系统,已成功应用于"2012年江苏·秦淮灯会"。

In previous crowd density estimation methods, an estimation model obtained by one camera cannot be applied to other cameras. In order to solve this problem, a method for cross-camera crowd densi- ty estimation was worked out. The proposed model mapped video data to a GIS coordinate reference sys- tem. So authors can process crowd images under geo-referenced coordinate system using GIS at same spa- tial scale. Test results show that one crowd estimation model obtained by a particular camera in GIS envi- ronment can be applied to other cameras. So cross-camera estimation model overcomes the dependence on surveillance scene and greatly improves the efficiency of crowd density estimation modeling. Finally, a sys- tem was designed and implemented for real-time monitoring regional crowds spatial-temporal distribution pattern based on cross-camera model, video surveillance system and GIS. This system has been successful- ly used in ＇ Jiangsu Qinhuai lantern festival, 2012＇. It can provide early warning information and a sound basis for safety and security decision making.