Tunnel Vision

Tunnel vision is generally considered a bad thing, unless you actually want to want to know what’s in the tunnel, in which case it’s a very good thing. This is a job that CCTV systems have handled for years, but they suffer from a big weakness: like any surveillance system, they rely on having people watch the monitors to spot any trouble.

Trouble is, it’s hard to remain vigilant when staring at a bank of screens, so the obvious move – obvious to me anyway – is to add machine vision tools to detect abnormal situations. But if you think about what goes on in road tunnels, that’s not so straightforward.  Cars and trucks come in many different shapes, sizes and colors, they move at different speeds, the spacing between them can vary greatly, they can overlap in the image, and so on. How then do you decide what constitutes a situation that needs intervention?

I don’t know, but I know a company that does.

Kapsch make an “Incident Detection System” for just such an application. A recent press release gives a very general overview, and also touches on the issue of calibration. I’m not entirely sure why calibration is needed, but it’s interesting that they do it.

More details of the capabilities of their tunnel monitoring systems are given in the Incident Detection System brochure, which is downloadable from  this page. It provides a fascinating insight into what tunnel vision can do for you.