Nowadays with the abundance of available video evidence from surveillance cameras and mobile phones, law enforcement agencies are using face recognition to reduce the time spent on manual viewing to find suspects.
Nowadays with the abundance of available video evidence from surveillance cameras and mobile phones, law enforcement agencies are using face recognition to reduce the time spent on manual viewing to find suspects.
“As the amount of video material submitted for investigations has significantly increased, investigators are looking for powerful tools to automate the search for subjects in videos,” said Elke Oberg, Marketing Manager at
Cognitec Systems.
“Face recognition helps law enforcement agents increase the speed and accuracy of criminal investigations by quickly identifying a face in photographs or in images from video footage in comparison to large image databases (e.g., mugshot repositories), and instantly receive a candidate list of possible matches,” she explained. “The software should be able to determine and cluster identities, and then display the appearances of the same person across a variety of media files.”
Elke Oberg,
Marketing Manager,
Cognitec Systems
It also allows police to act upon match results in the critical time period after a crime has been committed. “In general, using automated recognition engines for suspect searches saves investigation time and costs,” she added.
Automated searches, however, still require human involvement. “Face recognition is used as a supporting tool for law enforcement investigations, and match results are always evaluated by human experts. They are not used as permissible evidence in a court of law. Agents can set a low match threshold, and then use sufficient time and their expertise to look at the candidate images to determine a match,” Oberg explained.
Face recognition can also help police officers accurately evaluate and predict potential dangers that may arise, according to Roger Rodriguez, Retired NYPD Detective and Director of Client Relations at
Vigilant Solutions.
“For example, this might apply during an investigative stop when unknown subjects come in contact with law enforcement and have no valid forms of identification on them. It is also used as a check-in system for released prisoners in the probation space and can identify registered sex offenders to determine whether these subjects are adhering to the conditions of probation or supervised release,” Rodriguez said.
Law enforcement is also using face recognition technology for humanitarian reasons as well, Rodriguez explained: “Facial recognition technology can identify deceased persons who do not have valid forms of identification on them and to identify persons who cannot identify themselves, such as persons stricken with dementia, Alzheimer’s or amnesia. It is also often used to help identify and locate lost or missing children.”
Key requirements
Much of the controversy over law enforcement’s use of face recognition involves the technology’s accuracy, and wrongly identifying an innocent person. This is understandable considering law enforcement agencies primarily use face recognition technology for identity verification.
Compared to commercial applications, which often need more instant results, law enforcement demands high levels of certainty that a face match belongs to the right person. In order for law enforcement agencies to effectively and properly use face recognition technology there are several key requirements that should be met.
“Many law enforcement agencies seek facial recognition deployments which offer real-time video streaming and a forensic tool for post-event investigative analysis,” Rodriguez said.“Delivery models of these applications are offered on diverse platforms such as the web, PC, tablets and mobile devices.”
The ability to process videos at a speed faster than real time is also needed, according to Laura Blanc, Marketing Manager at Herta Security. “Our solution is able to process 24 hours of video within less than three hours. This makes forensic work much easier and faster. Being able to search for faces amongst large databases of millions of people without having false positives is key.”
Oberg added, “The matching algorithm should be trained with diverse image data, therefore preventing any bias effects, for neither race, gender nor age.” To further prevent wrong matches, she explained that investigators should be able to use tools to compare and inspect image pairs, such as line-blending, to find unmistakable facial features that can further substantiate the confidence in a match.
The use of cloud-based storage solutions could offer agencies a better means for collaboration and data sharing environments. “Systems with easy-to-use, intuitive pre-processing tools allow agencies to verify identities from images they could not otherwise have leveraged previously. It is an essential tool for law enforcement to generate investigative leads,” Rodriguez said.
Lastly, every law enforcement facial recognition deployment must have built-in metric tools which govern the oversight of the entire system. These reporting mechanisms provide agencies with necessary metrics while offering accountability and transparency.