Training Cameras to Track People: Opportunities and Concerns

By December 04, 2014

Researchers at the University of Washington, USA have developed a way of training sets of surveillance cameras to link up in order to track pedestrians closely. This of course raises a number of privacy and ethics questions.

Fixed surveillance cameras have been used for a long time now to spot intruders and/or record what is happening at a particular place. Now however they will be able to monitor people sequentially as they move around in a wider area. A team of electrical engineers at the University of Washington (UW) have developed an algorithm which enables cameras to attribute a colour and a number to individual people who show up, so that they can be recognised when they appear again. The novel aspect of their work is however that the monitoring process is not confined to a single camera’s field of vision: the UW programme allows a set of cameras to forward data on an individual to each other so that s/he can be tracked across their entire territory. In fact the movements of several people can be tracked simultaneously within the area surveyed by the cameras. Naturally this raises a number of issues relating to privacy and data security.

Improving urban security/providing marketing opportunities

There are two basic fields of application for this tracking system that immediately spring to mind. The first is in the world of law enforcement. Police forces could use it to spot and track people behaving suspiciously. If the UW system were combined with a system for detecting suspect behaviour, this would be a big step towards automating the location and tracking of criminal suspects, which could well make police work far easier, although the system does not in itself enable positive identification of a person by name.  However, given the sensitive nature of this kind of information, a clear legislative framework would need to be created, specifying who has the right to access such data. The second major field of application is in the retail sector, providing stores with a means of tracking customer behaviour in the shop.  Based on this, retailers could then set up systems designed to tailor offerings to the individual consumer, e.g. sending coupons/special offers to the customer’s smartphone or displaying on-screen reductions at the shelves where the customer appears to show most interest.

Smart City tool

Aside from the obvious commercial advantages of collecting large amounts of relevant sales data, the UW researchers envisage developing their system further – for example for the purposes of real-time mapping of pedestrian flows on a platform such as Google Earth. If sufficient numbers of surveillance cameras were installed, this could radically improve transport planning and also feed into in-car support systems designed to help motorists negotiate busy town streets and junctions.  This kind of  analysis of urban resident and traffic flows would constitute another step towards the Smart City. Whether such potential progress outweighs individual privacy concerns is a matter for debate but Jenq-Neng Hwang, a Professor of Electrical Engineering at the University of Washington, who is lead researcher on the project, stresses the positive argument: “Cameras and recording won’t go away. We might as well take advantage of that fact and extract more useful information for the benefit of the community,” he insists.

Legal mentions © L’Atelier BNP Paribas