Urban surveillance networks

I didnt know, until the terrorist bombing attacks, that most London buses have video cameras installed on them. There are thousands of cameras in London and increasing numbers in New York, Chicago and other major cities. Large swaths of the downtown areas are covered. Coverage goes up significantly if you include private cameras that monitor stores, parking lots and office buildings.
Camera surveillance networks have real benefits crime and traffic fatalities go down, and they generate useful evidence for a post-facto investigation but the limitations are obvious thousands of hours of tape that look like Warhols Empire State (Warhol pointed a camera at the building for 8 hours when a pigeon flew by at hour six, audiences burst into applause since it was the first thing to happen).
the key to better surveillance is to replace human watchers with computers. Once the imagery has been translated into bits, software can look for patterns has that car circled us twice, how did that pile of trash get to the roadside - and can merge imagery with data from other sensors (infrared or sniffers). Some call this intelligent video surveillance. Londons transit system already links fixed cameras to an Automatic Number Plate Recognition (ANPR) computer system to identify cars that park or drive in bus lanes.
Using urban surveillance networks to prevent attacks (rather than to prosecute the attackers after the fact) is in the too-hard category for now. Some prototype systems will notify an operator when the network detects a suspicious pattern, but this works best when tracking cars rather than people. A lot more code would need to be written to make urban sensor networks able to warn in advance of a mass transit attack. This is the false negative problem the attacker walks by the camera without triggering an alert. So where we are now is that a city could deploy a sensor network but it couldnt make use of the data generated for early warning and prevention of attacks. Putting lots of cameras on subway lines might have a deterrent effect, but my guess is that this would be minimal for suicide bombers.
The usual concerns are (1) privacy and (2) false positives, where a system would incorrectly flag a face or a behavior pattern as suspicious. Some people worry about the use of this technology for political control, and the place where this seems to be happening is (surprise) China, where the Golden Shield project includes constructing a digital surveillance network in Chinas cities.
Here are a few links: http://dtsn.darpa.mil/ixo/ (for battlefield applications); http://www.sarnoff.com/products_services/government_solutions/homeland_security/index.asp (critical infrastructure protection); http://www.tfl.gov.uk/tfl/cclondon/cc_fact_sheet_enforcement.shtml
Posted by Jim Lewis