Face recognition software is still in its infancy, and as such, some issues still remain with accuracy and ability. Systems have actually gotten quite good at methods of recognition under strict variable control; problems arise when a human face is under a shadow or obscured even slightly. Of course, these situations (shadows, partial images, etc.) are when facial recognition software would be at its most valuable, so it’s important to find ways in which to enhance the technology. As a leader in light measurement instrumentation, we are sharing how illumination variances improve facial recognition.
With that in mind, researchers have unveiled a fresh approach to face recognition technology that uses light measurement instrumentation to cut through hazy illumination and still accurately identify a person. Said researchers, based at Toyohashi University of Technology in Aichi, Japan, use an extended reflectance model that processes faces no matter what visual lighting obstacle may be causing obstruction in the first place. Named OptiFuzz, the system uses an algorithm to run a large array of illumination ratios (or possibilities) and systematically focus in on areas of obscurity.
In this way, according to lead professor Jun Miura, “By just adding this contrast adjustment to present face recognition systems, we can largely improve the accuracy and performance of face detection and recognition.” These improvements look to filter out the negative results that come with the effect of contrasted light and make face recognition software more agile and precise.
These new methods were run through a gamut of tests, including Viola-Jones Face Detector and the Mutual Subspace Method. These tests proved largely successful, so much so that researchers believe that their new methods could become useful under even the harshest illumination situations. And it’s important to note that even with advances in face recognition software, currently, techniques are only in place that work under full and consistent facial illumination.
Careful calibration of the contrast adjustment is key to success within the OptiFuzz system, and furthermore, light measurement instrumentation must be equally to task in calibrating and maintaining such a system. The system is still in an experimental phase, and even though scientists at the Toyohashi University of Technology are confident enough to release these findings, suitable systems for recognition replication will have to become codified. It’s here that a company steeped in light calibration technology can be of particular use, and Gooch & Housego finds itself uniquely positioned and excited for further developments in light contrast face recognition systems. For more information, contact the leader in light measurement instrumentation, Gooch & Housego, at (800) 899-3171.