Processing surveillance images - the growth opportunity for the next generation

from mwillett.org

Now that I've got your attention, just what am I on about today? Today's discussion is not necessarily about Photoshop; but jobs in the business of processing surveillance images.

Sir Richard Branson, founder of the Virgin Group, believes a sad reality of big business is people tend to forget decisions have to be made. It is Sir Richard’s opinion that this lack of decision-making is the main reason so many large companies suffer from inertia. What does this have to do with today's discussion? CCTV cameras. Yes, CCTV cameras.

CHICAGO -- More than 2,000 surveillance cameras in public places would be tied in to a network that would use sophisticated software to spot emergencies or suspicious behavior under a plan announced Thursday by Mayor Richard Daley. "Cameras are the equivalent of hundreds of sets of eyes. They are the next best thing to having police officers stationed at every potential trouble spot," Daley said.

WASHINGTON -- The Department of Homeland Security is funneling millions of dollars to local governments nationwide for purchasing high-tech video camera networks, accelerating the rise of a "surveillance society" in which the sense of freedom that stems from being anonymous in public will be lost, privacy rights advocates warn.

In the last month, cities that have moved forward on plans for surveillance networks financed by the Homeland Security Department include St. Paul, which got a $1.2 million grant for 60 cameras for downtown; Madison, Wis., which is buying a 32-camera network with a $388,000 grant; and Pittsburgh, which is adding 83 cameras to its downtown with a $2.58 million grant.

Recent examples include Liberty, Kan. (population 95), which accepted a federal grant to install a $5,000 G2 Sentinel camera in its park, and Scottsbluff, Neb. (population 14,000), where police used a $180,000 Homeland Security Department grant to purchase four closed-circuit digital cameras and two monitors, a system originally designed for Times Square in New York City.

"We certainly wouldn't have been able to purchase this system without those funds," police Captain Brian Wasson told the Scottsbluff Star-Herald.

Other large cities and small towns have also joined in since 2003. Federal money is helping New York, Baltimore, and Chicago build massive surveillance systems that may also link thousands of privately owned security cameras. Boston has installed about 500 cameras in the MBTA system, funded in part with homeland security funds.

By the time governments around the world are finished (if that ever happens), billions will be spent on building up a massive surveillance infrastructure. The amount of data and images produced by this network will boggle the mind. And here is where Sir Richard's point is made; we are spending tons of cash on the hardware, but ignoring the fact that there needs to be processing done on the back end. The very thing that you gain from CCTV, an immediate picture to put on the evening news, is lost when agencies either have no staff or minimal staff to process these images for the media and for court. Here's a typical example:

Let's say a crime is committed in an average store in an average town in the US. The store recently bought a CCTV surveillance system from Costco in order to get an insurance discount. It was a great deal, a 4 camera system for under $500. It'll pay for itself in the first year as their insurance carrier discounts their policy $800 per year for having the system installed.

The crime is committed and the police are called. The store owner mentions that he has CCTV and has watched the footage and he can identify the person on the video. Small problem though; he can't figure out how to get the video out of the system. When he bought the system, he never thought about getting the video out should something happen. (click on the Costco link above and review the specs of the system before continuing. This is a popular system in our area and represents a good example of what the average person is buying and installing in their place of business)

The detectives arrive and ask the owner for the video. He tells them that he can't figure out how to get the video out of the system. This leaves the detectives with a series of choices.

1. Watch the video on the screen and take notes; but leave the system behind.
2. Watch the video on the screen and take notes; then seize the system as evidence.
   
3. Watch the video on the screen and take notes; then call "their computer guy."
   
4. Watch the video on the screen and take notes; then call in a specialist from private industry.
   
5. Some combination of the above.

Let's say, for the sake of argument, that the detectives correctly process the DVR as evidence and deliver it to their department's forensic video analyst (if their department has one). What will the analyst most often find? For this model, the display resolution is 720 x 480 (PAL=720 x 576) but the

record resolution has two options, 320 x 112 or 640 x 224. The DVR was set to the default of 320x112 and about 7 frames per second. The DVR uses "Modified Motion JPEG" compression (modified=proprietary) and each frame is about 12k. (320x112=35840 - 35840 does not equal 12k and what did the manufacturer throw away with their modified compression?) (a recent surf found a 5MP camera from Fuji priced under $100)

What do you think the chances are that a successful identification can be made?

But wait, there's more ...

If you looked carefully at the specs, you'll notice that there is no included option for digital output, like a CD or DVD burner. The system more than likely uses an embedded version of the Linux operating system. That hard drive won't exactly play nicely when you attempt to attach it to your PC (please don't try this unless you've been trained and have the correct tools). So your only option is to go out the CVBS port to a analog recorder; right? Look at the specs. The display resolution is 720x480 ... and the recorded resolution is 320x112. Ouch.

Going from 320 to 720 involves an increase of 225%. Going from 112 to 480 is not so easy to do in your head and it is definitely not 225%. How does the DVR do it? I don't know. You won't know. Chances are, if you can get through to the Australian company that makes this DVR, they won't be able to tell you either. So what do you do, other than encouraging them to buy a better system to replace the dodgy one that now sits in evidence?

If forensics is debate and being able to explain what you've done and why it was the appropriate choice, you're really in for it here. But this is not the point that I am trying to make.

Sir Richard's point

In our example, dozens (if not hundreds - depending on the number of images) of hours are spent on the video from this one case. If the agency has a forensic video analyst, there are still only so many hours available in his/her work day. Governments and private citizens are making the conscious decision to deploy CCTV for numerous reasons. They are not making the decision, however, to support the eventual processing of those images. What does this have to do with inertia ("a body in motion tends to remain in motion, a body at rest tends to remain at rest"). Many folks tend to refer to inertia as an amount of resistance to change in velocity. Newton aside, government and industry are at rest, and are tending to stay that way (for the moment).

The amount of pressure building behind the collective back-log of unprocessed images is enormous. Some agencies are behind as much as 6 months to a year. Some agencies are refusing to process images from petty crimes. Some larger agencies help those smaller agencies without any technicians; but only on murder cases. Is justice being served?

Gen-X and those younger are inheriting this debt of unprocessed images. We are also looking at a tremendous growth opportunity. I see ads in Craigslist and on the photographer sites for folks looking for work. Here's the work ... processing surveillance images for court and for the media, trying to get something usable from that 320x112 image, developing a better DVR, developing that next piece of software that does this all automatically, and so on.

If you are not in the work already, where do you start? Colleges like the British Columbia Institute of Technology, private training from groups like LEVA, or simply start reading books like George's new Photoshop book for forensic folks (and hopefully you read mine when it comes out as well). These are all great places to start. Others will come along as well as the pressure continues to build. 

So, to the new generations, the opportunity is ours to take.