For today's column, I visited the Johns Hopkins University security office in Remington (nicely situated across from Dizzy Izzy's and Charm City Cakes) to get an update on the SECURES gunshot detection system they've installed.

In short, sensors positioned around Charles Village and other neighborhoods register gunshots and can almost immediately pinpoint their location (to within 10 feet), allowing police to respond quickly. Police in Washington use a similar program and have expressed delight. City cops are still looking at the system and aren't sure yet. Hopkins is a good test, though I've expressed concern the sensors are in low-crime neighborhoods (the university got it for free, so it's hard to criticize them for putting up on their own turf).

But while in the security office, I was far more intrigued by another system Hopkins uses -- image recognition software. It was fascinating to watch this in action.

A dispatcher sits in front of six computer screens, many divided with up to six different camera locations, and a large screen on the wall that shows even more live locations. With 155 cameras spread over the campus and adjacent neighborhoods, it's impossible for the two dispatchers to both keep an eye on everything much less notice when something goes wrong.

That's where the image recognition software comes in. The head of Hopkins security, Edmund G. Skrodzki, a retired Secret Service agent, showed me an alley in which a student had recently been mugged. A camera is now positioned to record who goes in and out. It would be tedious for one person to keep watching the alley 24 hours a day, much less unproductive, so they've programmed the computer to send an alert whenever a person enters. On a screen next to the dispatcher, that image then flashes on and the person watching knows to pay close attention. This is set up at bike racks and parking garage entrances as well. A yellow box even surrounds the image on the screen, and the dispatcher can decide -- in the case I saw, it was obviously a student -- whether to send a patrol car out.

Nearly 100 bikes were stolen last year, before the software was installed. This year, only three have been reported missing.