US Military Scientists Create Facial Recognition System for Use in the Dark

Researchers at the US Army Research Laboratory have developed an algorithm capable of recognizing faces from thermal imaging. The preprint of their study was published on arXiv.

The developers trained a convolutional neural network to match images of people’s faces taken with a thermal imager to regular photographs. On an infrared image, the neural network identifies the outline of the face as well as key features such as the mouth, nose, and eyes. Using this data, it creates a synthesized image and then searches a database for the closest match.

Tests have shown that the program already works with 85% accuracy.

Potential Military Applications

The US Army Research Laboratory’s press service stated that this technology will be used during nighttime military operations. It will help identify especially dangerous terrorists or high-ranking individuals in the dark. The program can be used in real time by connecting it to a thermal imager and a computer, or after an operation is completed.

According to The Next Web, the new recognition algorithm could be used in combination with another US military development—a thermal imager capable of seeing through walls.