Back to list of 2004/05 seminars

Abstract for Seminar

The ability to collect pedestrian flow information in real time, without the need for gathering and post-processing low-level data to extract measurements (such as density and flow rate), is a goal that up to now has proved infeasible. There are a number of reasons for this: the volume of data, the cost of processing, the abilities of the people observing the scene or subsequent video tapes and the effects of variation in the lighting conditions of the area being observed.

I shall describe low cost infrared sensors that can be used to track the movement of pedestrians within their fields of view. The resulting data streams (30 frames per second) are processed in real time to generate density, flow-rate and speed data and person counts every two seconds. This information is displayed and also saved in a file. In addition, the paths taken by all pedestrians can also be shown in real time and saved for playback and further processing.

I shall describe an experiment designed to test the ability of the system to cope with large numbers of people moving in a simple space, with counter flows and an entrance to one side. The current limitations of the approach will be described and I will show how these can be overcome to construct a scalable system that can be used to monitor pedestrian movements in any size and shape of pedestrian space.

The concurrency mechanisms in the system are entirely built on CSP design primitives and implemented using Quickstone's networked extensions to Kent's JCSP concurrency packages for Java ... and will be discussed!

Computer Science @ Kent

	Last modified Thursday September 23 15:20:29 BST 2004 Problems with this page? Contact the CS Webmaster
http://www.cs.kent.ac.uk/dept_info/seminars/2004_05/abs_2004_10_05.html