South Bend, Ind. — University of Notre Dame researchers have developed a new face-matching technique that may become an effective law enforcement tool to help identify criminals who have undergone plastic surgery to alter their looks.
A report in New Scientist notes that existing facial-recognition software can be fooled even by minor changes in the lighting and position of an unaltered face. New Scientist quotes Notre Dame computer scientist Kevin Bowyer, Ph.D., who headed the project, as saying that surgical changes of one or more parts of the face reduce by half the success of existing software in matching “before” and “after” photos.
Investigators theorized that matching individual features, rather than whole faces, could be more successful for identification purposes. Their inspiration was a technique known as “sparse representation,” which matches an image of a face by comparing it with combinations of individual features from faces recorded in a database. If the closest matching combination comprises features primarily drawn from one person in the database, it’s likely the target image is also of that person. If the best match combines features of many different people, then the system has failed to identify the new face.
The new system uses two databases: a general one containing random faces, and another containing all of the “before” pictures — much like police mug shots. When a target “after" picture is analyzed, a composite picture as similar as possible is created from the features of images in the general database. All “before” pictures go through the same process, according to New Scientist. If the composite picture created from the “after” picture matches closely with any of the composite pictures derived from the “before” pictures, the two are declared a match.
Researchers found that while surgery changes the appearance of a face, many individual features remain the same, and matching based on the nose or eyes alone was more accurate than matching based on some existing whole-face techniques. Combining the matches of all facial features gave the team a 78 percent success rate when comparing pre- and postsurgical photos.
A spokesman for a company that provides facial ID software to policy says that while the new approach can help law enforcement officials identify surgically disguised criminals, it is unlikely to ever be totally accurate.
Go back to the Cosmetic Surgery Times eNews newsletter.