Composite artistry meets facial recognition technology : exploring the use of facial recognition technology to identify composite images

Abstract

Forensic art has been used for decades as a tool for law enforcement. When crime witnesses can provide a suspect description, an artist can create a composite drawing in hopes that a member of the public will recognize the subject. In cases where a suspect is captured on film, that photograph can be submitted into a facial recognition program for comparison with millions of possible matches, offering abundant opportunities to identify the suspect. Because composite images are reliant on a chance opportunity for a member of the public to see and recognize the subject depicted, they are unable to leverage the robust number of comparative opportunities associated with facial recognition programs. This research investigates the efficacy of combining composite forensic artistry with facial recognition technology to create a viable investigative tool to identify suspects, as well as better informing artists and program creators on how to improve the success of merging these technologies. This research ultimately reveals that while facial recognition programs can recognize composite renderings, they cannot achieve a level of accuracy that is useful to investigators. It also suggests opportunities to better design facial recognition programs to be more successful in the identification of composite images.