The Scoop on NJIT's Facial Recognition Security System
Just beyond the automatic sliding doors of the Robert W. Van Houten Library, you’ll find shelves teeming with books, dissertations and literary journals, a front desk manned by friendly librarians—and an unassuming yet fully-functional facial recognition security system currently safeguarding the throng of NJIT students who frequent the four-story, modern research facility.
It’s called FacePro: a server-based analytics technology, which provides facial search and matching capabilities on real-time video streams captured by four Panasonic 6 series 1080p i-PRO cameras that hang in the corners of the entryway of the library (two facing in; two facing out) near two flat screen monitors.
As people walk through the field of view of the cameras, the system captures images of faces and juxtaposes them against those stored in a database, which is processed and kept on a server and managed using two workstations housed in the FacePro Lab inside the library.
Remember when the idea of facial recognition seemed like some far-flung pipe dream contrived for sci-fi movies? Well, it is plenty real—and ever evolving.
But pioneering technology is only part of the story here.
“In 2014 we had 11 thefts in the library, which is higher than normal,” says Robert Gjini, director of security systems at NJIT. The uptick in property theft—laptops, mobile devices—set Gjini and his team on a crusade to tighten up security by finding a video camera surveillance system with monitors that displayed people being recorded as they enter and exit the library in the hopes that it would deter stealing.
Coincidentally, around the same time, Panasonic reached out to NJIT about an innocuous live-camera face matching software that they were looking to bring to the United States.
“It was the perfect match at the perfect time,” recalls Gjini. “We had the security need. They wanted to set up shop at a beta site, and they thought NJIT would be a good place to do it. It worked out really well.”
FacePro contains two databases: One is populated with some 70,000 images of people—students, faculty, alumni—who are allowed on campus. (The photos come from the Security Systems, Photo ID and Parking Services Department). The other database catalogs people who are banned from the NJIT campus for a number of reasons.
“When the system has a match with someone who appears on camera with someone who isn’t allowed on campus, an alarm is generated and Public Safety is called and will respond,” explains Gjini.
The security systems team at NJIT was assisted by Panasonic with the initial set up of the pilot program and implementation of FacePro and continues to work alongside the NJIT Department of Public Safety to keep the photos of people who are barred from campus current and up-to-date.
Since FacePro was installed Feb. 20, there haven’t been any library thefts reported on campus.
News of the radical decline in theft comes at a fitting time, especially during the end of the semester, when the library, which is open until 2:45 a.m., becomes a circadian enclave for students and study groups cramming for finals and finishing up end-of-the-year projects.
FacePro offers a bevy of innovative features. More than a simple facial detection system, it’s also a multifunctional, technologically savvy platform capable of recognizing facial features and discerning specific demographics down to skin color, age and gender.
Some of the images uploaded to the database of people prohibited from campus are provided directly by the Department of Public Safety. Once uploaded, dots are superimposed onto the picture to detect eye position before it is evaluated and registered as a faceprint.
FacePro also offers a slew of high-speed search capabilities, allowing the user to play live and recorded video and sift through thousands of imported photos—by day and time—taken by the high definition cameras. There’s also an option to search the general database to look up the latest triggered alarms and view a statistics report that provides analytical information based on facial recognition data.
This recognition “Robocop” of sorts seems to be the wave of the future, as the line between science and fiction continues to thin at a rapid pace.
Last summer, Facebook’s artificial intelligence team announced that its facial recognition software passed key tests with near human-level accuracy.
And the FBI has Next Generation Identification in the works, which will serve as an incremental replacement of the Integrated Automated Fingerprint Identification System, providing advanced lookup capabilities, including palm print, iris and facial identification.
Testing on a cloud network in a beta environment allows NJIT to reap the safety benefits of the technology while reps from Panasonic, who visit the campus weekly, can test the system, pull information and adjust the settings to ensure effectiveness.
“These type of systems are very prone to false positive alarms,” says Gjini. “Let’s say that John Smith is in the system as ‘do not allow’ and somebody walks through who looks like John Smith; the system does occasionally give a false positive hit. Panasonic plays with the algorithms and adjusts the sensitivity of the lighting schemes at the entrance to get the false positives as low as possible.”
Given the success of the pilot program, it looks like FacePro has found a new home at NJIT—and is here to stay. In fact, there’s even talk of expanding to other high traffic areas, like the campus center.
“With the help of our partners at Panasonic, FacePro has proven to be a sound choice to help overcome some security challenges, identify persons of interests on our college campus and drastically lower incidents of theft in the library,” says Gjini. “It’s fitting that this modern, cutting-edge security system would play a crucial role in upholding the value we place on safety at NJIT, where sophisticated technology permeates all aspects of student life.”
By Shydale James