We learned this after the Boston Marathon bombing, when Boston Police Commissioner Edward Davis told The Washington Post that the department’s facial recognition system hadn’t been able to identify the bombers, despite the fact that both had official driver’s license and immigrant records in the system.
If recognizing faces is problematic, the search for objects is no easier. “We want it to identify an object left behind in a crowded environment, but as much as we would like to be able to solve that problem, differentiating a bag left behind versus an airport patron standing two feet away from their bag — the false alarm rate is too high for effective management,” said Warren Brown, president of video technology provider ObjectVideo.
That’s not to say that facial recognition and object identification remain the stuff of science fiction. Various software products are available today that can conduct these tasks with varying degrees of accuracy. As with all technology, though, the real usefulness often is demonstrated not by the power of the tools, but by how they are used.
A video camera system is only as good as the sum of its parts.
“When people talk about camera, they think about a camera as a single entity, but the reality is it’s a network of systems, and the components on that system need to work together,” said Sony’s Collett. Sony has focused its efforts on image quality, which is a crucial element in any recognition algorithm. But those images still must be paired with powerful processing software, and few standards exist to make those pairings happen consistently.
More significantly, the camera and the software together don’t think like a law enforcement professional. In fact they don’t think at all. In order for the system to go the last mile, someone must program in the parameters. At what point, under what circumstances, will a real-time surveillance system send out an alarm?
“The software isn’t artificial intelligence,” Collett said. “You have to program the system to actually notify you of the conditions that you are worried about. This means that someone has to define those various conditions and define what actions need to occur by humans, and what actions need to be taken by other systems that may be integrated.”
Here the burden may well fall to the emergency management community. As the professionals most closely attuned to the broad spectrum of possible risk, “they can create the set of rules and conditions in their environment,” Collett said.
This begins not just with the analytics process, but also with the initial systems deployment. “It’s not just about picking a camera that may look good at 10 o’clock in the morning when the sun is shining to the left,” Collett said. “What about when the sun is shining straight overhead? It’s not just about megapixels. People who are making the selections need to think through the environment in which the camera is going to be in use.”
In return for these efforts, emergency managers may reap substantial rewards from enhanced video surveillance. A more thorough video feed to the command center, for instance, could enhance a manager’s ability to monitor a situation and respond to changes.
It could also stream video to first responders’ laptops as they head to the scene of an incident. “It can give them valuable insight into what they are going to be confronting once they get there,” said Rob Sprecher, North American practice director for public safety and justice at Unisys.
As technology improves, it will become increasingly possible to pair this broad situational awareness with specific details — a license plate number, for example — that could cue in first responders to the possible risk factors in a situation.
To reap those rewards, emergency managers will first have to make a compelling case for having cameras on the scene, and that may not be easy.
“Emergency management ultimately reports to the citizens of the community, and there will be some citizens who are absolutely against it, some who will be for it, and some who will be for it but with limitations,” Hudson said. Suppose you want to put a pan-tilt-zoom camera on a street corner. That isn’t going to happen until you convince the neighbor across the way that the camera won’t be looking in her window.
Civil liberty advocates have been raising objections to the vision of a public environment in which every person and every action is observed and recorded. For some, London’s pervasive ring of video surveillance looks deeply Orwellian.
These fears only grow worse with the proposition that drone aircraft should be used for surveillance. Boston’s police commissioner has said he wants unmanned aerial vehicles for next year’s marathon.With the public already perceiving drones in primarily a military context, the use of unmanned aerial vehicles for video surveillance is raising hackles.
Some see potential for compromise. The public agreed to give up some liberties when airport metal detectors proved they could curb terrorism. It’s possible that successes by video surveillance could similarly sway public opinion in favor of more video.
“The debate has a lot to do with who has access to those feeds, and who is watching those watchmen,” Young said. “If you lack faith in the government’s credibility to watch out for your interests, the more you suspect whether this is a power that is reasonable for government to have.”
Here again it may be up to the emergency management community to steer the ship.
“What are you capturing, who has access to it and what will you do with it?” Brown said. “If the only place this is being used is in the first responder community, and the only time they can access it is when they are addressing a crisis, then there are ways to manage that issue.”
This story was originally published by Emergency Management magazine.