Knowing the true identity of an individual who gives false information, or no information at all, is a problem faced by many in law enforcement. Did the person in question leave his wallet on his dresser like he claims, or is he wanted on one or more active warrants? Is the prisoner being released from the jail really who he says he is, or has he assumed another, similar-looking prisoner’s identity? The answer can mean the difference between life and death for the officer or deputy.
Until recently, there were few ways to determine who an individual really was. While motor vehicle records could be searched if a photo driver’s license was presented, the identity of a person without any identification, or a driver’s license that had been altered, was difficult to determine. To combat these unknowns and save lives, the Pinellas County, FL Sheriff’s Office is one of the first agency in the country to use groundbreaking facial recognition technology in a mobile setting.What is Facial Recognition Technology?
Facial recognition technology makes use of unalterable features of a face, such as the distance between the centers of the pupils of the eye. It then uses an algorithm, a finite set of steps for solving a problem, which converts the image to numbers. The software program is then able to compare the digital photograph of a face with others in the database and bring up matches in gallery format, with the most likely match first. The officer or deputy then decides which of the matching images, if any, are of the person in question.
The Pinellas County Sheriff’s Office (PCSO) uses software developed by Viisage
of Billerica, MA. The company developed the software specific to the needs of the sheriff’s office, using their data and incorporating the fields they needed such as name, address, and prior arrests. This software, when used in conjunction with digital cameras and laptop computers, has allowed the sheriff’s office to identify wanted persons as well as to prove that individuals are telling the truth about their identities. In 2004, Viisage acquired ZN Vision Technologies, a European leader in facial recognition technology, making the system even more efficient and accurate.Using Facial Recognition Technology
PCSO has installed the Mobile Identification System (MIS) software as well as digital cameras and docking stations in 50 of its 600 cruisers, going live in June 2004. Now, when deputies come across someone whose identity is in doubt, they take three to four pictures of the individual. They then put the camera in the docking station connected to the mounted laptop in the cruiser and extract the photos by pressing a single button. Taking several photos at the same time allows the deputy to select the best one to be scanned and eliminate those where the person has turned his head, his clothes have gotten in the way, his eyes are closed, or the photo itself is blurry.
Once the pictures are downloaded and the deputy highlights the one he wants to use, the computer uses wireless communication to search the Pinellas County database of 850,000 records. The computer brings up the 24 closest matches in the order of greatest probability. The deputy can then decide at his mobile location if the person really is who he says he is.
Currently, the deputies can also search the databases of five other large county sheriff’s offices in Florida including Miami Dade and nearby Hillsborough County, as well as the Florida Department of Corrections. Access to Broward County’s database is being worked on. Right now deputies can access three million images, and with the addition of Broward County, there will be about four million images.
Deputies generally search the Pinellas County database first because 90% of the people are found there. If the person is not in that database, they then search neighboring Hillsborough County; however, they can search all records in the entire database if the need arises.
When the deputy feels there is a match, information about the individual appears on the screen of the laptop. This information has been tailored to the Pinellas County Sheriff’s Office’s specifications and includes such things as date of birth, number of arrests and why, outstanding warrants, and when the photo was taken. A separate system is used to check for warrants once the correct name is established.
It is up to the deputy to discover if there is a match. Because PCSO began taking digital images in 1994, some of the images in the database are over 10 years old. So while the technology gives the deputy a head start, he is the one who must decide if the pictures chosen by the computer include a match.
The system can also be used to confirm true stories. One man had the same middle initial, last name, and date of birth as someone who was wanted. He also had tattoos. Only the first names were different. When his photo was run through the system, a match was found (because of a previous arrest). The positive identification was made due to the photographs and descriptions of the tattoos. He was not the wanted person.
The use of mobile facial recognition in the cruisers is just one part of a bigger system used by the sheriff’s office. In 1994 PCSO began replacing the existing mug shot procedures with digital photographs. Thus, when the facial recognition technology was incorporated, they already had digital images to use. (They did not convert old hard copies to digital because doing so would be time consuming and the photos would be out of date).
Now at the intake facility, photographs of each individual are taken three different times. The first is taken when the new inmate comes through the door and the resulting photograph is downloaded into the system. If the person had previous contact with law enforcement and shows up on the screen, the staff can know such things as whether he was previously violent, or whether he should be allowed to bail out. This first photograph is also used to establish who the individual is, his physical condition upon arrival at the jail, and any jewelry or accessories he might have.
The second photograph is a standard mug shot series with glasses on and off, profile view, and pictures of any marks, scars, or tattoos. This is taken when the person is admitted to the jail. Other text information is also added to his file, such at date of birth, and saved for future investigations.
The third photograph is taken at the release desk to verify that this really is the person to be released—this is a one-to-one search. The resulting picture is also entered into the system because appearances change, especially after a jail term.
All of these photographs are immediately available to all law enforcement agencies in Pinellas County and also statewide. In addition, the Criminal Justice Network (CJNet), a networking infrastructure administered by the Florida Department of Law Enforcement that allows Florida’s criminal justice agencies to share information at all levels of government, is also linked back to PCSO.
In Investigation—Facial recognition technology can also aid in the investigation of crimes. Unfortunately, many businesses use low-resolution surveillance cameras that satisfy the very minimum requirements of the insurance company, and the pictures they take may not be useful. However, some produce better quality photographs that can be put into the system. In fact, any digital JPG photograph can be entered and a search initiated. A digital photograph can even be taken of a driver’s license photo and matched to the database. This could be used to discover that the photograph is not the one belonging on that license.
The system could also be used to identify a body, again if the person has had previous contact with law enforcement. If a good photo can be taken of the deceased, he may be successfully identified. However, if the face has been changed due to decomposition, facial recognition technology cannot be used.
Facial recognition technology is also being used at the St. Petersburg/Clearwater airport, at the Pinellas Courthouse, and jail visitation center, to check a small local database of active warrants.Cost, Financing, and Equipment
In 2000, PCSO received funding from a federal grant through the Office of Community Oriented Policing Services (COPS) administered by the US Department of Justice. The goal of the grant was to demonstrate the use of facial recognition technology by Florida law enforcement. The sheriff’s office began with the booking process at the Pinellas County intake facility, and later expanded to the Mobile Identification System (MIS).
The initial cost was about $250,000. Most of this outlay was for the design and development of the software for a system that, up until that time, didn’t exist. Expansion of the MIS to include more cruisers will involve mainly the cost of the additional cameras and docking stations.
The cameras used by MIS are off-the-shelf Hewlett Packard Photosmart 850s with standard docking stations. They were chosen because they can be easily updated and are simple to use. Features include four megapixel resolution, auto focus and flash, zoom lens, and output in JPG format.
The camera was also selected because of the docking station. While most docks are more or less flat, the HP Photosmart 8881 Camera Dock has sides to hold the camera securely so that it won’t fall off due to the movement of the cruiser. The docking station also allows for one-touch connections to the laptop for image downloading, and recharges the camera’s batteries when it is docked.
It was found that the vehicle’s vibration was starting to wear down the pins on the docking stations, leading to poor connections. To correct this and also ensure even more stability, the department added a strap to hold the camera in place. They took one of the docking stations apart and located spots where the frame had no electronic components behind it. They then fastened one end of the unused, rubber cord-wrap straps that came with their Dell computers to the docking station near where the lens of the camera would be. Holes—already in the strap—slip over a screw that protrudes on the other side.
The cameras are mounted in their docking stations next to the laptops in the cruisers to make them easily accessible, with the lens facing forward, away from the car’s interior. The commander selects the cruisers receiving facial recognition technology. Deputies with a large amount of contact with the general public, such as traffic detail, use these vehicles. The cruisers with the technology are also dispersed throughout the shifts and patrol areas in the county.Training
Little training was needed for the deputies to take good pictures with the camera because of its intuitive nature. However, while facial recognition software is a giant step in aiding law enforcement in discovering the true identity of an individual, there is still the human element when interpreting the data it presents.
Learning to use facial recognition technology includes classroom training. Four to eight deputies at a time receive a four-hour block of education about the system and its use and participate in practical exercises. These include searching through the image galleries to identify and select known individuals from a photograph presented.
Sometimes the database photo is of poor quality or 10 years old, so they learn to look at unchanging facial features such as the eyes, eyebrows, or shape of the mouth or nose. They need to be able to pick out a person from a group of photographs whether or not he has grown a beard, shaved his head, or gained weight. Sometimes there are drastic changes in the way an individual looks from one picture to the next.
Deputies also need to check the written information that appears with the image, being aware that while people may change their date of birth or name, these changes are generally small ones. They might change the year but not the date and month of their birth to make it easier for them to remember, or they might change a couple of numerals in their social security number.Obstacles and Problems
A major obstacle that must be overcome when incorporating facial recognition technology is that of unrealistic expectations. Nothing in life works as perfectly as it does on TV, yet we have come to expect that machines can tell us with 100% certainty whether the person is, indeed, who he says he is. To combat that, PCSO taped several excerpts from television shows that might lead deputies or the general public to have these expectations.
In one excerpt, the lead characters started with a photograph of the upper side of a person’s face. It included one of his eyes; the rest of his face was covered with a gas mask. The show’s crime lab computer matched it with a photograph of the back of someone’s head with 100% certainty. This just doesn’t happen in real life.
Officers and deputies need to be trained in the intended use of the system. They must learn what the technology can and, more importantly, cannot do. The deputy is still the primary decision-maker; the system is just a tool to produce likely matches. He must go through the images and decide which the correct one is.
On the legal side, the deputy still needs probable cause. He can’t arrest an individual just because of what the facial recognition system says. If the person is not under arrest, he must give consent before his picture can be taken. He has the option of saying no, turning away, or covering his face.The Future
PCSO would like to equip more cars with this technology once funding can be found. It is currently installed in a little less than 10% of the marked cruisers. They would also like to incorporate more counties into the database. Right now they have access to only seven out of 67 Florida counties. And they would like to link to the federal database, as well as share information with other law enforcement agencies.
Because of the Mobile Identification System, in the first nine months of use PCSO has made 53 arrests that otherwise would not have been possible. And of those arrests, about half were felony warrants. Since the system has been in place in the corrections environment, not a single person has been mistakenly released.
According to Lieutenant Jim Main, Project Director/Technical Services Division, there are two certainties in police work: people want to fight, and people lie about their names. The implementation of the TASER has helped solve the problem of fighting, and now facial recognition is helping to solve the problem of lying.Jim Weiss is a retired lieutenant from the Brook Park, OH Police Department and a frequent contributor to LAW and ORDER.
Mickey Davis is a Florida-based journalist.