Smart Cameras Blog

[From FBI says facial recognition not ready for prime time | Homeland Security News Wire]

James Loudermilk II, a senior level technologist at the FBI, said “the agency sees no point in facial recognition.” I believe part of the problem is that facial recognition doesn’t work as effectively as other biometric technologies for the FBI.

Facial recognition would be the killer application of biometrics, Loudermilk told the hundreds of conference delegates, and the FBI would dearly love to be able to use facial recognition in its fight against crime. It can not, though. The algorithms just do not exist to deliver the highly reliable verification required. This is even though the FBI has been evaluating facial recognition technology since 1963, he said. It did not invest then. It is not investing now.

Instead, the future strategy for the FBI focused on iris prints and existing fingerprint and DNA databases. In another presentation, Loudermilk noted several other biometric measures. These included “facial, retinal, population comparison studies, voice patterns, etc.”, with the use of near real time matching.

As an example, consider how the Integrated Automated Fingerprint Identification System (IAFIS) operates. Currently, 18,000 law enforcement agencies contribute fingerprints and DNA samples to the FBI’s databases and, at their peak, they submit 200,000+ identity verification queries a day.   IAFIS database contains the fingerprints and corresponding criminal history information for more than 55 million people.

This older presentation offers some interesting statistics on the IAFIS system. The average response time is 15 minutes in FY07 and 98% are done within 2 hours. IAFIS processed 26.1 million submissions in FY07 (44% Criminal, 56% Civil) and 92% are received electronically. 85% of fingerprint checks handled without human review. Facial recognition can’t claim the same level of efficiency.

rshah Facial Recognition

[From ShotSpotter Hits the Suburbs as the Police Fight Gun Violence - NYTimes.com]

An article on ShotSpotter mentioned an independent study on gunshot detection.

In 2008, Peter Scharf, a criminologist at Tulane University in New Orleans, conducted a study for the National Institute of Justice about an early competitor of ShotSpotter, called Secures, in two Virginia cities. He found the system frequently sent the police on wild goose chases by reporting false positives, had an inconclusive effect on response time and, crucially, had little impact on arrest rates.

The report is very well done. Some of the interesting findings include how the gunshot detection system was negatively affected by radio frequency interference and fireworks/bottle rockets.

The report also discusses the tradeoffs between false positive and false negative error (an unavoidable part of these systems). The report found that “2/3 of SECURES®-related dispatches were “but-for false alarms” – both not a confirmed gunshot and no call corresponding call for service.

rshah Gunshot Detection

[From Chicago's red-light cameras don't always deter accidents -- chicagotribune.com]

I am hoping to review this data in a more detailed post later. But for now, here are some important snippets:

Cameras are said to reduce accidents, but collision records compiled by the Illinois Department of Transportation indicate that accidents increased at many city intersections the year after red-light cameras were installed. In fact slightly more intersections saw an increase than a decrease, the data show. The city tells a very different story. Crash statistics compiled by the city reflect broad success in reducing accidents with cameras, and the city could not explain why the numbers are so different.

The read from the state numbers is this: Although some Chicago intersections indeed appear to benefit from the presence of cameras, nearly 60 percent do not.

By year’s end red-light cameras will be installed at 189 Chicago intersections, the most of any big U.S. city. Sprawling Los Angeles, where the car is king, has 32; New York, 150.

This work casts doubt on the city’s claim that the cameras reduce accidents. (If the cameras don’t reduce accidents, then they aren’t providing the promised safety benefits.) Read the city’s claims here:

The City of Chicago reports crashes have been reduced by 20 percent in the two years since the camera technology was installed in early 2006 at 10 intersections. . . . Crashes decreased 30 percent, from 1,055 in 2004 to 736 last year, at intersections where red-light cameras were installed in 2004 and 2006, according to the Chicago Office of Emergency Management and Communications.

rshah Chicago, Red Light Cameras

An interesting academic paper is “War Rooms” of the Street by Torin Monahan. The abstract:

This article investigates the surveillance dimensions of “intelligent transportation systems” in the United States, with a particular focus on the mediation of data by engineers in transportation control centers. These communication systems lend themselves to surveillance by means of “function creep” beyond their primary intended purposes and through the everyday collection and manipulation of data to manage mobilities. In the U.S., dominant system protocols privilege vehicular throughput and discipline those who deviate from that norm.

I found some interesting insights. First, these intelligent transportation agencies are very secretive. It is difficult to gain access to them (as an academic) and there is little written about their capabilities. Second, Torin discusses the significant surveillance capabilities these agencies wield. Obviously, it is intended that the surveillance be used for transportation and emergency needs. But Torin is able to tease out some examples of function creep, where these surveillance powers are used for other purposes, such as police work. A good read for people interested in transportation systems and the ever growing creep of surveillance.

rshah Policy

[From YouTube - Video Analytics in Chicago (video)]

A news video from ABC 7 (full text) that provides an overview of the chicago camera network. Demonstrates a couple of analytic techniques such as setting a spatial trip wire, identifying left bags, and identifying cars in traffics. Otherwise no new info.

rshah Chicago

[From Illinois Lawmakers Push Red Light Camera Restrictions from the Newspaper.com]

Three bills that would limit red light cameras:

The most ambitious proposal comes from state Senator Dan Duffy (R-Barrington) who last month introduced Senate Bill 2466 to wipe any authorization of automated ticketing machines from the statute books.

State Representative Angelo Saviano (R-Elmwood Park) has a more limited objective in mind. Last month he introduced legislation that would not ban cameras entirely, but instead would ban the most lucrative form of ticketing. “A county or municipality, including a home rule county or municipality, may not use an automated traffic law enforcement system to provide recorded images of a motor vehicle for the purpose of issuing violations to persons driving a motor vehicle who enter an intersection to turn right, or to turn left from a one-way street into a one-way street,” House Bill 4631 states.

State Senator Rickey R. Hendon (D-Chicago) offered the least ambitious of the red light camera measures. His bill would prevent localities from issuing a ticket for a certain type of technical infraction. “A county or municipality, including a home rule county or municipality, may not use an automated traffic law enforcement system to provide recorded images of a motor vehicle for the purpose of issuing violations to persons driving a motor vehicle who come to a stop one foot or less past the point where a driver is required to stop,” Senate Bill 2477 states.

While these bills have an uphill battle, it is good that some legislators are willing to stand for a principle against easy money.

rshah Chicago, Red Light Cameras

[From Chicago's Camera Network Is Everywhere - WSJ.com]

A good overview article of the current state of Chicago camera network. I was interviewed for the article. Couple of snippets:

While many police forces are boosting video monitoring, video-surveillance experts believe Chicago has gone further than any other U.S. city in merging computer and video technology to police the streets. The networked system is also unusual because of its scope and the integration of nonpolice cameras.

The city links the 1,500 cameras that police have placed in trouble spots with thousands more—police won’t say how many—that have been installed by other government agencies and the private sector in city buses, businesses, public schools, subway stations, housing projects and elsewhere. Even home owners can contribute camera feeds.

Rajiv Shah, an adjunct professor at the University of Illinois at Chicago who has studied the issue, estimates that 15,000 cameras have been connected in what the city calls Operation Virtual Shield, its fiber-optic video-network loop.

Former U.S. Homeland Security chief Michael Chertoff has called Chicago’s use of cameras “a model for the country.”

rshah Chicago

[From Atlanta seeks to add 500 surveillance cameras | ajc.com]

The Urban Institute is conducting a study on the surveillance cameras in Chicago. Here is an article with their preliminary results:

Preliminary results from the study show crime has fallen in parts of both cities where the cameras are located. For example, violent crime and larcenies fell by 25 percent — or 30 incidents per month — in downtown Baltimore, starting in the fourth month after the cameras were installed in 2006, the study says. The cameras, according to the study, have helped identify suspects and getaway cars and find weapons used to commit crimes. “It has helped solve literally thousands of crimes,” Chicago police spokesman Roderick Drew said. “In fact, our detectives have reviewed over 20,000 video segments this year alone.”

The Urban Institute’s study, meanwhile, has found the cameras are not without their problems. When they automatically pan areas, they may capture only portions of a sequence of events. At night and during bad weather, they might not capture images strong enough for evidence. They are sometimes targeted by vandals. And their maintenance costs can be high.

I will be looking forward to the final study and its associated data. I don’t fully understand the results stated in the article. It took 4 months before crime fell? Maybe it was something other than the cameras?

rshah Chicago, Policy