A Train Station Surveillance System: Challenges and Solutions

In this paper, we describe our multi-year development effort to create a smart-camera surveillance system for use in train stations. We address the problem of activity analysis in very crowded areas and offer solutions to the detection/tracking problems in one of the most crowded environments. The smart camera system is installed in several train stations around Tokyo, Japan and is currently deployed analyzing crowd movement and individual gestures of the passengers in the stations. The paper summarizes the technical problems we faced during our research /development stages as well as during system deployment. The hierarchical smart camera system, which can detect the overall crowd movement in the higher level and individual activities in the lower level, is developed by Verificon Corp., a Princeton University startup company. We developed smart camera algorithms and spent several years commercializing the technology in a joint project with Yokogawa Electric. A railway company was the lead customer on this project. Verificon system passed several rounds of testing, running for several months on live data in several Tokyo train stations and was deployed by the railway company. This project has led to several extensions of the original technology.

[1]  Wayne H. Wolf,et al.  Human detection in compressed domain , 2001, Proceedings 2001 International Conference on Image Processing (Cat. No.01CH37205).

[2]  Jason Schlessman,et al.  Hardware/Software Co-Design of an FPGA-based Embedded Tracking System , 2006, 2006 Conference on Computer Vision and Pattern Recognition Workshop (CVPRW'06).

[3]  Mubarak Shah,et al.  A Multiview Approach to Tracking People in Crowded Scenes Using a Planar Homography Constraint , 2006, ECCV.

[4]  Mubarak Shah,et al.  Floor Fields for Tracking in High Density Crowd Scenes , 2008, ECCV.

[5]  Wayne Wolf,et al.  Design of a Real-Time Gesture Recognition System , 2005 .

[6]  Takeo Kanade,et al.  Tracking in unstructured crowded scenes , 2009, 2009 IEEE 12th International Conference on Computer Vision.

[7]  Wayne H. Wolf,et al.  A peer-to-peer architecture for distributed real-time gesture recognition , 2004, 2004 IEEE International Conference on Multimedia and Expo (ICME) (IEEE Cat. No.04TH8763).