Researchers develop image-enhancement technology for facial biometrics in dark and over distance
Researchers with the Science and Technology division (DST) of Australia’s Department of Defense have developed thermal technology to enhance biometric facial recognition algorithms in order to identify individuals in adverse conditions, such as at distances of up to 250 meters, according to the DST website.
DST biometrics research team members Sau Yee Yiu and Dmitri Kamenetsky are co-authors of a paper on “Image Enhancement for Face Recognition in Adverse Environments,” which also describes facial recognition working in environments such as a dark alley on a moonless night. The new algorithms created by the team were successful at matching subjects even across large distances and with near-zero visibility.
Yiu says the direction of research was focused by a literature review in the early stages.
“I then came up with a model of how heat propagates through the atmosphere, and this turns out to be similar to the way noise from atmospheric turbulence distorts images over long distances,” she says. “The atmosphere moves and shifts around and your image gets sheared and blurry. Applying my heat dispersal model gets rid of that turbulence and brings it back closer to a focused, sharp image.”
Graininess is removed from low-light images by passing them through various filters. The researchers also developed an interface to control several parameters with sliders to interactively tweak the algorithm. The output is updated in real-time, enabling the user to tailor the algorithm to the particular environment.
A separate algorithm used to judge the overall quality of facial images showed that the images processed with the modified algorithm were superior quality than the originals, consistent with visual inspection, according to the researchers.
The team presented its results at the 2018 Digital Image Computing: Techniques and Applications (DICTA) conference.
“We’re very happy with the results, which will be of benefit to stand-off surveillance systems,” says Kamenetsky. “We’ve released a description of the algorithm, allowing other researchers to implement it and make further improvements. Interestingly, most of the research presented at DICTA was using deep learning in some way, ours is just a relatively simple yet effective mathematical approach.”
Researchers have previously developed technologies for ultra long-distance imaging, but it is not clear the images could be effectively used with biometric facial recognition. OmniVision has also developed a sensor for ultra-low-light conditions that can perform facial recognition in darkness at the edge.