In the realm of security and search and rescue operations, color image-based systems are integral in the location of people but with current technology, not adept at distinguishing actual human skin with other objects that possess similar hues. In an attempt to develop a solution to this problem, researchers from the Air Force Institute of Technology have created a two-dimensional feature space with a heightened ability to characterize human skin due to its use of the spectral absorption characteristics of melanin, hemoglobin and water, as outlined in their study.
The system can focus in on specific aspects of humans tissue by using a skin index that takes into consideration how water and melanin manifest at different wavelengths in the near-infrared region. Using this system, the overall cost of hyperspectral-based research and rescue systems could potentially by reduced by a factor of seven.
"The study represents a crossroads between physics and statistical pattern recognition," Michael Mendenhall, co-author of the study, said in a press release. "The features were designed based on an understanding of the physics behind skin's spectral shape, but in such a way that the features separated skin and non-skin pixels in order to make the pattern recognition portion of the problem more effective."
Current image recognition programs are typically too slow for the intense nature of search and rescue operations and due to the high cost of many hyperspectral systems - around $700,000 - many organizations are unable to afford them.
Mendenhall and his team plan on continuing their research in this field by focusing on the properties of human hair in order to give devices the ability to better recognize mixes of skin and hair and account for the ability of human skin to reflect light in a mirror-like manner.
