The attempt by law enforcement authorities in Boston to use crowdsourced video images and still photos to identify those responsible for the fatal bombings in the city earlier this week could prove challenging, but it's not unprecedented.
In 2011, the city of Vancouver, British Columbia, resorted to a similar tactic to identify the main participants in a riot in the city following a loss by the Vancouver Canucks to the Boston Bruins in Game 7 of the Stanley Cup finals. The riot resulted in more than 100 injuries and $4.5 million in damages to vehicles and businesses.
There are probably hundreds of people with footage showing the person or persons who left the bomb. Grant Fredericks, video analysis instructor, Law Enforcement and Emergency Services Video Association
To identify the ringleaders, the Vancouver Police Department's riot investigation team gathered more than 1,600 hours of video images and nearly 1 million digital images in dozens of different formats from the public, the media and closed-circuit TV cameras in the city.
A team of 50 forensic analysts from the Law Enforcement and Emergency Services Video Association (LEVA) in Indianapolis, helped the Vancouver police analyze the images. Working out of a multimedia evidence-processing lab at the University of Indianapolis, the team of analysts first converted the video images into a standard format and loaded them into centralized video servers from where the images could be accessed by several people.
Working around the clock over a two-week period, the analysts reviewed the footage and added metadata and unique identification markers to describe individuals and specific events, such as the time or location where an incident might have occurred or the direction in which an individual might have been heading. More than 15,000 events were tagged as criminal acts using this process.
The infused metadata allowed investigators to search through the images and to track individual behavior and events between video footage from multiple sources, said Grant Fredericks, an instructor at LEVA and technical manager of the forensics effort. The analysis resulted in criminal charges brought against more than 200 individuals involved in the rioting.
The city's effort remains one of the largest video forensics efforts involving crowdsourced video and images.
The same approach can also be used in the Boston investigations, but the sheer volume of data will make the task harder, said Fredericks, who is also a video analysis instructor at the FBI National Academy in Quantico, Va., and the founder of Forensic Video Solutions.
"When you have an incident like Monday in Boston where you have 150,000 to 200,000 people -- a significant percentage of those people have cell phones and they are taking pictures and video," Fredericks said, "Every single one of them was potentially collecting evidence for this investigation," Fredericks said. "There are probably hundreds of people with footage showing the person or persons who left the bomb," without their knowing it, he said.
Law enforcement authorities investigating the Boston Marathon bombings on Tuesday called on citizens to turn in any video footage or digital images they might have taken at the event. According to Boston Police Commissioner Ed Davis, investigators intend to go through "every frame of every video" in their possession to track down the perpetrators of the attacks that killed three people and injured more than 170 others.
The immediate task for Boston investigators will be to ensure that they have a good process in place for gathering video and digital images turned in by the public, Fredericks said. In Vancouver, many of those who sent in video of the riots edited the footage or changed file formats while submitting their videos and compromised the integrity of the footage in the process, he said.
The next task will be to convert or to transcode all the images they receive from the public and from the hundreds of CCTVs in the area of the bombings, into a standardized interoperable and uncompressed format, Fredericks said.
In the Vancouver investigation, analysts at LEVA used a software tool called Omnivore from Ocean Systems, a Burtonsville, Md.-based company, to convert the video from hundreds of different screen resolutions, screen sizes, frame rates and analog and digital file formats into a standard format.
The standardized footage will then need to be stabilized to ensure consistency in quality, said Charles Guarino, vice president of Ocean Systems. Once the images have been stabilized, investigators looking through the video will have the ability to tag or mark anything significant they see in the footage.
"It's like putting a yellow sticky tab on the video," he said. The tabs allow investigators to instantly get back to a specific point in the video footage using specific search terms.
Such tagging will allow investigators to capture sequential images from different video sources and track individuals who have moved between different video sources within a specific time period, Guarino said. "You could line up multiple sources of video and sync them all up," and see events as they happened in sequential order even though the images might have been captured by multiple video devices.
Many forensics video tools these days also have audio acquisition and processing capabilities that enable video searches using specific words or phrases. "I could take a section of video and type in a word that you might have said and [the software] will phonetically search the audio and video and cue me right up to the place where you said 'bomb'," Guarino said.
"In Boston, the challenge is going to be getting all the data they can," said Paul Steinberg chief technology officer at Motorola Solutions. The company acts as an integrator with police department and governments in the U.S. and elsewhere to set up surveillance systems that are used mostly to review an incident after the fact.
"Some images might be already posted on social networks, so there's lots of places it can come from and lots of differences in the quality of the data at different resolutions on video on smartphones and handheld cameras," he said. Generally, the coarser the resolution, the less precise the result will be, he said.
"In the past there have been cases where software was used to correlate faces and abandoned objects. Software is even used to detect the amount of time a person spends in front of a display," Steinberg said. As the general quality of cameras goes up, the usability of the forensics tools increase as well, he said. " So there's much more there now to use."
Senior Editor Matt Hamblen contributed to this story.
Jaikumar Vijayan covers data security and privacy issues, financial services security and e-voting for Computerworld. Follow Jaikumar on Twitter at @jaivijayan or subscribe to Jaikumar's RSS feed. His e-mail address is email@example.com.
Read more about security in Computerworld's Security Topic Center.