Facial recognition, video forensics bolster human expertise

October 5, 2015 - 

This is a guest post by Robert Sprecher, Program Director for Face Recognition and Video Analytics for MorphoTrak.

Thanks to rapid advances in facial recognition and video forensics, criminals and terrorists are finding that technology is increasing the likelihood of their rapid apprehension. Video forensics, which is the review of video evidence, and especially the emerging science of facial recognition are undergoing accelerated development in application and effectiveness. Although both technologies have been available in some form for decades, their usefulness is increasing as software, hardware, biometrics, and related fields of scientific study mature.

Ronald Carnes, 68 years old, was recently arrested in Iowa while attempting to license a vehicle and apply for a driver’s license. Unfortunately for Carnes, and fortunately for the general public, Iowa’s Department of Transportation uses facial recognition software, which led to the identification of Carnes not as the person whose name he was using, but as a 1973 prison escapee who had been convicted of armed robbery in North Carolina.

Understanding how the technology works is essential to appreciating its power. Facial recognition, much like fingerprint identification, works on mathematical “plotting” of points from recorded images, whether they are fingerprint ridges or facial characteristics such as the corners of the eyes. By leveraging these features from an image on an algorithm, the “face” can be compared with others in a database. If a possible match is found, investigators have narrowed their search immensely.

The technology does have limits. Although identifications may take only seconds on a television show like CSI, even automated systems can be much less speedy, and search results (“candidates”) require verification to confirm a match. Other variables mean the process is not as quick or as certain as it is usually depicted on TV. Face comparison and verification of matches are performed not by machine, but by trained examiners who evaluate the candidate images submitted by the face recognition software.

The Iowa driver’s license case is a good example. That catch occurred under nearly ideal circumstances. Driver’s license applicants are required to stand in a certain way and are told not to smile. These and other steps are designed to reduce variables as much as possible, making the photos as “facial recognition-friendly” as possible. So, when Mr. Carnes posed for his driver license in Iowa, the odds were already in favor of finding a match with his North Carolina criminal photo booking record. Without these quality control standards, it is possible Mr. Carnes would still be at large rather than in a North Carolina jail anticipating completion of a 17-year-sentence.

It’s interesting to note that facial recognition software was not immediately used as an investigative tool in two tragic, high-profile bombings – the Boston Marathon bombing on April 15, 2013 and the London Underground bombing on July 7, 2005. (Later, a variety of facial recognition tools were used to support the investigation.) In the early absence of automated video analysis and facial recognition software, investigators developed the suspects by manually reviewing long hours of video and large amounts of photos, looking for anyone carrying a bag or wearing a backpack.

Once investigators localized all people with bags and backpacks, they began tracing the individuals’ movements. Applying these manual methods of forensic video analysis, the initial suspect was derived when an agent reviewing store-front video noticed that the behavior of the suspect’s companion was opposite from almost everyone else at the blast. This observation, along with other factors, led to the identification of the suspects as the Tsarnaev brothers.

In the case of the London bombing, authorities isolated images of individuals with backpacks and bags at the scene immediately before the blast and then traced them backwards, all the way into the far suburbs, by their clothing and bags. Manual face comparison was used to further pinpoint the suspects.

These cases illustrate the dedication and expertise of the law enforcement professionals who worked tirelessly to capture the terrorists. And yet, the cases also suggest how much faster and more efficiently investigations could occur through the use of automated video analytics, and face recognition. Certainly, video analytics will never replace investigative techniques used by trained experts. But as with fingerprints, video analytics and face recognition are rapidly providing additional tools for law enforcement, security professionals, and military personnel who must always work to stay ahead of the bad guys.

DISCLAIMER: BiometricUpdate.com blogs are submitted content. The views expressed in this blog are that of the author, and don’t necessarily reflect the views of BiometricUpdate.com.

Leave a Comment

comments

About Rob Sprecher

Mr. Robert Sprecher PMP is the Program Director for Face Recognition and Video Analytics for MorphoTrak. Prior to this role, Mr. Sprecher was Practice Director for Unisys North American Public Safety and Justice Program. He has more than 30 years of national and international law enforcement experience, in addition to technology development, architectural programming, and project management for federal, state, and local agencies. He worked as a Deputy Sheriff, Detective, SWAT Commander and Department Director for large, urban public safety agencies. He received his undergraduate degree from Regis University and attended the University of Colorado Graduate School of Public Affairs.