John Honovich points out the importance of false positives. A false positive arises when a smart camera system warns operators of a situation, when the situation is normal. This problem arises when cameras move out of laboratories into real world conditions where weather and external stimuli, such as birds, make video analytics difficult.
Here are some interesting snippets:
Making sure you only trigger when someone is violating is hard because there are so many factors that might set off the analytic besides a valid adversary. To a computer, rain, dust and snow can all generate a similar form to a human being. Quick changes in light or the movement of water (waves) can also generate such forms. A camera that shakes because of the wind or issues with the mounting or installation also can trigger such alerts. The hard part in such analytics is to make sure that these alerts can be eliminated. This is a key metric in testing and differentiating between analytics.
False positives drive up the cost of systems. . . . While it is better operationally to centrally manage alerts, if the system generates dozens or hundreds of false alerts per day, the costs can become prohibitive. Let’s say an ‘intelligent’ camera generates 5 false alerts a day at the cost of $1 per alert (the unit cost to pay a monitor to assess). That’s $5 per day, over $1,800 per year and about $10,000 for a projected 5 year lifecycle. If you dozens or hundreds of cameras, this hidden operational cost can be in the millions. And this is not theoretical. This is the feedback you will hear time and again from real world deployments. It’s widely accepted that this is improving but it is still the major factor in duing your due diligence in analytics.
Developed by Karl Martin and Kostas Plataniotis, researchers at the faculty of engineering, their secure visual object coding application uses cryptography techniques to encrypt “objects of interest” within video frames -B faces or other features that may be used to identify a person – and store them separately. In order to view the original complete image, a decryption key is needed to restore the object of interest.
The project web site by Karl Martin shows a number of pictures of the process at work. The idea here is to encrypt either faces or complete bodies as a way to protect privacy. For example, surveillance could be monitored in real-time but faces would be obscured. However, this is reversible if a decryption key is used.
The PET application is an add-on designed to work with a front-end video surveillance system that has object detection capabilities. “Facial detection has been around for a long time to varying degrees of performance. Many vendors already offer it in consumer applications like digital cameras so you can focus on faces when taking photos.”
They are currently working on implementing their system with the Toronto Transit Commission (TTC).
An interesting article on the use of small cheap motion detectors within buildings. They can be used as an alternative to cameras in some areas, because they are cheaper and less invasive from a privacy standpoint. It is suggested that this technique could be used for security monitoring.
Here is a nice video of the system, Ambient Intelligence for Better Buildings (link to Journal article)
Thruvision offers a camera that relies upon Terahertz electromagnetic waves. The promise of terahertz is that these waves emit from people, but pass through solid materials. By passively collecting them, it is possible to identify if someone is concealing an object. The above story has lots of links for more information.
A nice article on the growing use of facial recognition. A couple of interesting points:
After a driver sits for a photo at the Illinois Secretary of State office to renew a license, officials use facial-recognition technology to give the resulting image a close look. First, state officials verify that the face matches the images portrayed on previous licenses issued under the driver’s name. The second, more extensive run-through determines if the same face appears on other Illinois driver’s licenses with different names.
Since starting the program in 1999, the state has uncovered more than 5,000 cases of multiple identity fraud, said Beth Langen, policy and program division administrator at the Illinois Secretary of State office. The state pays Digimarc Corp. about 25 cents per license for the service, she said. . . . About 40 percent of the nation’s drivers are set to undergo such facial-recognition database checks when they renew their licenses in 20 states. . . . .
Police departments, eager for more investigative tools, are pressing for access to the millions of photographs in the motor vehicle databases. A few states prohibit such sharing, but many allow it.
Gait analysis and recognition is another useful application of smart camera systems. There are several problems with reliable gait analysis including the collection of data and that the choice of footwear can affect how people walk. Plantiga offers a new shoe that can provide more gait and movement information and then transmit it wirelessly. As a result, it offers a unique method of biometric identification. For more, see this page on their web site.
I have no idea how well this works, but it is a novel and interesting approach.
Smoke and fire detection is another application for smart cameras. A nice introductory video by Fire Sentry is online at youtube. For more background, see these academic articles, 1 and 2. A quick search also found this a camera with embedded smoke detection capability by axonX, known as SigniFire.
DARPA is supporting various smart camera technologies in a number of ways. Here are snippets from the story:
Dynamic Multisensor Exploitation, or “DYME,” aims to combine cameras, radar, and acoustic sensors to better find bad guys as they move through urban canyons, and along coastal waters.
The goal of “Building Labels for Urban Environments,” or “BLUE” (you gotta love these acronyms), is to automatically label the structures seen in surveillance video.
TheCombat Video Analysis Engine(no acronym, alas) would use “computer vision, machine learning and probabilistic models to detect and recognize complex threats and suspicious activities without identification of specific individuals.”
I was looking up the latest on facial recognition and I found a few interesting sites and stories.
First, the Metro Nashville, TN school system is installing face-recognition technology, which is suppose to be operational by December 2007. Its a bit strange that this is installed in an elementary school, but I will be awaiting the results of the test.
Second, a facial recognition research group is online at http://www.face-rec.org/newsgroup/. There are lots of techie articles, but also an active google group (face-rec) discussing the issues.
Third, the National Institute of Standards and Technology has been evaluating and testing facial recognition technologies. The project is known as FRVT. Its too bad there haven’t been other public tests for other features of smart cameras.
Great article, here are some interesting snippets:
Liberty Island’s video cameras all feed into a computer system. The park doesn’t disclose details, but fully equipped, the system is capable of running software that analyzes the imagery and automatically alerts human overseers to any suspicious events. The software can spot when somebody abandons a bag or backpack. It has the ability to discern between ferryboats, which are allowed to approach the island, and private vessels, which are not. And it can count bodies, detecting if somebody is trying to stay on the island after closing, or assessing when people are grouped too tightly together, which might indicate a fight or gang activity. “
My heist had been condoned by Pedro Ramos, Pathmark’s vice president of loss prevention, though he didn’t know precisely when or where I was going to attempt it. The beer was identified by an object-recognition scanner at ankle level—a LaneHawk, manufactured by Evolution Robotics—which prompted the cashier’s question. Overhead, a camera recorded the incident and an alert was triggered in Ramos’s office miles away on Staten Island. He immediately pulled up digital video and later relayed what he saw. “You concealed a 12-pack of Coronas on the bottom of the cart by strategically placing newspaper circulars so as to obstruct the view of the cashier.” (Using Storevision)
But examples abound. Take E-ZPass. Drivers signed up for the system to speed up toll collection. But 11 states now supply E-ZPass records—when and where a toll was paid, and by whom—in response to court orders in criminal cases. Seven of those states provide information in civil cases such as divorce, proving, for instance, that a husband who claimed he was at a meeting in Pennsylvania was actually heading to his lover’s house in New Jersey. (New York divorce lawyer Jacalyn Barnett has called E-ZPass the “easy way to show you took the offramp to adultery.”)
Recent Comments