The accuracy of facial recognition technologies decreases as a subject ages, according to new research from Michigan State University.
The MSU researchers sought to assess how effectively contemporary facial recognition systems can match a face that has aged over an extended period, testing the technology against mug shot images for about 23,600 individuals. The researchers found that for a six-year gap between images of the same individual, facial recognition systems tended to operate with 99 percent accuracy, but that beyond the six-year mark, “recognition accuracy begins to drop”, according to an MSU report.
The finds were published in the IEEE Transactions on Pattern Analysis & Machine Intelligence journal, and while the specific technologies used and the specific figures for accuracy decreases are only available to journal subscribers, the research offers a compelling pedigree, having been led by renowned biometrics expert Anil Jain, together with doctoral student Lacey Best-Rowden. With Jain being an expert sought after by police agencies and private companies alike, his new warnings that the facial biometrics of a given subject ought to be updated every few years are likely to be heard throughout the academic, governmental, and private sector communities concerned.
Source: MSU Today
March 3, 2017 – by Alex Perala