Application Note: IR Imaging Differentiates Skin from Other Materials
SANKT AUGUSTIN, Germany, Dec. 11, 2015 — Systems using IR wavelengths to distinguish skin from other materials hold promise for improving search and rescue efforts and defending against efforts to deceive biometric systems.
The proliferation of face-recognizing biometric systems for security — in settings ranging from public buildings and transportation to secure data centers — has, in turn, fueled efforts to circumvent or outwit these systems. Conventional imaging systems can be deceived by masks or other facial reproductions because they verify faces by comparing a reference image to a captured image, which can be falsified.
Results from and design schematics of an active multispectral shortwave-infrared camera system for skin detection and face verification. Courtesy of Holger Steiner/Bonn-Rhein-Sieg University of Applied Sciences.
Now researchers at the Bonn-Rhein-Sieg University of Applied Sciences have found a way to differentiate between human skin and other materials while concurrently checking biometric characteristics.
Skin can't be reliably identified at visible wavelengths due to the vast and nuanced variation of people's skin types and colors. Poor lighting conditions can also make identification more difficult. Even albedo — a surface's reflection coefficient — cannot be relied upon to clearly recognize skin in the visible range, as the values of skin types vary so much, and other materials such as meat, leather or wood can produce very similar albedos.
But the shortwave-infrared (SWIR) wavelength range (780 to 1400 nm) offers an opportunity. Above 900 nm, the researchers found, skin pigments have no relevant influence on albedo, thus absorption by the water contained in the skin increasingly gains in influence.
The investigation was implemented with a SWIR camera mounted in the middle of a ring of LEDs. Three rows of LEDs transmitted light in different wavelengths within the defined range onto the face to be analyzed. The camera recorded the reflected SWIR light, and the data was made available on a connected computer using the Vimba software development kit from Allied Vision Technologies GmbH.
Software written by the university then took over analysis, displaying the skin as brown, irrespective of type or color. Other materials appeared, each according to its consistency, in black or white, and could be explicitly differentiated from skin.
If facial recognition systems were equipped with such camera systems, they would authenticate users only if facial characteristics were correct and recognized as skin. LED illumination allows the system to be used independently of environmental lighting, and since SWIR light is invisible to the user, he or she would notice nothing of the scan.
Another research project at Bonn-Rhein-Sieg dealt with the use of skin detection methods in workplaces with high levels of work-process automation. Person recognition based on image capture was used to differentiate between objects and skin, to identify materials and to determine danger zones. When a human penetrated the danger zone with his or her hands, the robot or machine was decelerated or halted.
The researchers used the Goldeye G-032 SWIR camera from Allied Vision in both projects.
The work was published in the Journal of Sensors (doi: 10.1155/2016/9682453 [open access]).
Meanwhile, at the Air Force Institute of Technology (AFIT) in Dayton, Ohio, researchers are using near-infrared (NIR) imaging to overcome a challenge in aerial search and rescue missions: telling human skin apart from objects of similar color that cause false alarms.
Usually the image acquisition and processing required by hyperspectral imaging systems are too slow for live search and rescue operations. Furthermore, specific air platform requirements and the high cost of equipment acquisition and management — around $700,000, according to AFIT — mean hyperspectral systems are beyond the budgets of search and rescue organizations.
The AFIT researchers developed an imaging system keyed to the presence in skin of water and melanin, manifested at two NIR wavelengths. Limiting the imaging system's scope cuts the overall cost of such hyperspectral-based search and rescue systems by a factor of seven, the researchers said.
"After a lot of investigation into spectral properties of false alarm sources, we arrived at a simple observation that skin is more red than green, due to the melanin in darker skin and oxygenated hemoglobin in lighter skin, whereas many of the false alarm sources were more green than red," said professor Michael Mendenhall.
The researchers used the skin detection and false alarm-suppression feature to design an application-specific optical system using three framing cameras. Their first breadboard system is about 12 × 12 × 6 in. Because their skin detection solution can be implemented with less expensive technology capable of live video frame rates, its total price tag would be around $100,000.
The research was published in Applied Optics (doi: 10.1364/AO.54.010559).
- The technology devoted to the analysis of unique biological characteristics such as voice patterns and fingerprint, retina, iris, and hand and face geometry to determine or authenticate the identity of an individual.
MORE FROM PHOTONICS MEDIA