Facial Recognition in Law Enforcement
Written by John Dowden   

MUCH HAS HAPPENED in the evolution of facial recognition technology through the years, and many in this field would argue that some of the most exciting developments have happened in just the past few years. 

But first, a quick review.

The predecessor technique to today’s facial recognition technology had its beginnings many decades ago in the 1960s. These first semi-automated facial recognition programs required a human to locate features, take measurements between various features, and compare the resulting figures against reference data. It was a largely manual process.

By 1990, mathematical formulas were applied to the process, and the software that later evolved into today’s top facial recognition products was born. Since that time, we’ve gone from comparing 2D images against one another, to creating 3D models and capturing images in real-time to select distinctive features on the face in question. We’re also now leveraging high-definition video using better and less expensive cameras, which are widely deployed by law enforcement in many major cities.

As a result, we’re seeing some impressive, high-profile uses of facial recognition software out in the real world.

For example, during the summer of 2017, the U.S. Department of Homeland Security’s Customs and Border Protection (CBP) division began accelerating the rollout of its Biometric Exit pilot program. Mandated by Congress after the attacks on September 11, 2001, the program is designed to identify and capture wanted fugitives and known or suspected terrorists trying to flee the country.

Facial recognition software plays a starring role in the execution of the CBP’s pilot program. Now being piloted at some of the nation’s busiest international airports, Biometric Exit requires international passengers on certain pre-designated flights to step up to a kiosk or pole-mounted camera and simultaneously scan their boarding pass. Each passenger’s facial features are compared against the manifest of photos on file, taken from existing records and pre-validated images on file, as well as a watch list containing known criminals or persons of interest. This check helps ensure that the passenger getting on the plane is the same person on both the scanned ticket and on the manifest.

Benefits, and How It Works

Why all the focus on facial recognition? For starters, facial recognition software is frictionless and requires less contact with a subject of interest. It’s also easy to operate with minimal training. The user need not be an expert in biometrics in order to operate the tool.

And finally, it’s convenient. Video and still image cameras are now ubiquitous. They’re on smart phones, tablets, in computers, mounted in ceilings, walls, on buildings everywhere, and light posts around us. The availability of facial imagery combined with the continual advancement of facial recognition technology and accuracy has created a viable disruption in the speed and means by which law enforcement conduct their investigations or organizations secure their facilities and campuses.

Moreover, if the ultimate goal in public safety and security is to save and protect lives and catch criminals before they can inflict harm on others, then face recognition has the potential to be a real game-changer.

Think about the recent tragedies at the Route 91 Harvest music festival in Las Vegas, and the Pulse nightclub shooting in Orlando. These were major shooter incidents involving major loss of life. If we had the proper information and advanced recognition technologies in place, there’s a chance—not a guarantee, but a chance—that these incidents could have been prevented or rendered much less deadly than they were. Logic tells us that the lawful and proper use of investigative tools combined with good police work more times than not equals safer communities.

So, how does it work?

As a refresher, once the software detects a face in a photo or face scan, the system goes to work determining the size, position, pose and unique characteristics of the face in question.

Every face has numerous, distinguishable landmarks. These are the different peaks and valleys that make up facial features. These landmarks are called nodal points. Each human face has approximately 80 nodal points. Some of the nodal points measured by the software include:

• Distance between the eyes
• Width of the nose
• Depth of the eye sockets
• The shape of the cheekbones
• The length of the jaw line

The system translates nodal-point measurements into a numerical code or set of numbers, called a faceprint, representing the features on a subject’s face that can be compared to faces in the database. A match is verified from the faceprint.

Contrary to one popular misconception, face recognition systems don’t actually store photographic or facial images. They only store faceprint templates or features derived from processing the images. This helps with speed and storage savings by eliminating the need to store terabytes of duplicate image data.

Today’s most advanced facial recognition systems have achieved incredible accuracy and speed. The fastest systems boast an accuracy rate in the range of 98 percent or more, even with low image quality.

In order to reach these performance milestones, face recognition systems must go through rigorous testing by the National Institute of Standards and Technology (NIST), which is part of the U.S. Department of Commerce. The most recent of these tests relevant to facial recognition is the agency’s Face Recognition Vendor Test (FRVT) 2013 and Face in Video Evaluation (FIVE).

It’s probably no surprise that the highest performers in NIST testing tend to be the most widely deployed advanced recognition systems today.

Investigative Use Cases for Facial Recognition

The current most popular, plausible and actionable use for this technology is to help solve crimes that have already occurred, thereby bringing closure to victims and the families and friends of victims of violent and non-violent crime.

As a result, police departments are deploying facial recognition systems in their investigation-operation centers, where operators can collect and process image and video data from crime scenes around the community.

The following are a few examples of how the technology has been deployed with a good deal of success:

Chelsea Bombing—One of the most well-known examples happened in mid-September 2016, when explosions from three homemade explosive devices rocked neighborhoods in New York and New Jersey, injuring 31 people. It became known as the Chelsea bombing because two of the bombs were in the Chelsea neighborhood of Manhattan. Officers there found a fourth, unexploded device in Chelsea, where the first explosion had injured the vast majority of victims in the incident. Luckily, nobody was killed. However, for a few days the New York metropolitan area was on edge.

Within hours of the bombings, investigators identified and later apprehended 28-year-old Ahmad Khan Rahami as their chief suspect by matching his face from surveillance video to his image in a U.S. immigration photo database. Rahami was convicted by a jury in October 2017.

South Wales Police (United Kingdom)—Police in South Wales, United Kingdom, have deployed facial recognition systems in a mobile unit, which is a van outfitted with cameras and computers to locate and identify suspects while on-the-go. Initially piloted during a championship football match last year, the technology has enabled police to positively identify more than 190 suspects since that time. In December 2017, the technology was even reportedly used to identify the body of a deceased person.

City of Irving, Texas—The Irving Police Department has deployed facial recognition in its crime information center to support officers out in the field who are often investigating crimes with very little to go on, outside of video surveillance imagery. The tool can give officers in the field access to real-time information about suspects they may encounter in the field to make decisions and hopefully solve crimes faster.

Arizona Department of Transportation (ADOT)—In 2015, the Arizona Department of Transportation began using facial recognition technology to protect residents against identity theft and fraud. The process begins when a resident applies for or renews their driver’s license. Each photo taken at the DMV is checked against the existing database of photos to check for duplicates under a different name. Positive matches are sent to FBI-trained technicians to closely verify that both photos are of the same person. A detective then investigates further and collects all available information on the case to determine whether criminal charges should be filed. To date, more than 100 cases have gone to court.


About the Author

John Dowden is the Senior Product Manager for NEC Corporation of America’s Advanced Recognition Systems division. He has over 20 years of industry experience planning, developing and implementing multi-modal biometrics products and solutions for implementation and operation across the world. Before working in biometrics, he was both a military officer with the Air Force and electrical engineer within private industry.


This article appeared in the Summer 2018 issue of Evidence Technology Magazine.

 
< Prev   Next >






Recovering Latent Fingerprints from Cadavers

IN A HOMICIDE CASE, the recovery of latent impressions from a body is just one more step that should be taken in the process of completing a thorough search. This article is directed at crime-scene technicians and the supervisors who support and direct evidence-recovery operations both in the field and in the controlled settings of the medical examiner’s office or the morgue under the coroner’s direction.

Read more...