U.S. develops remote face recognition system to achieve target recognition within 1 km

The U.S. military is developing a portable face recognition device that can identify targets a kilometer away, according to a report published in The New Science. This is the “Advanced Tactical Face Recognition of Remote Technology” project being implemented by the U.S. Special Operations Command, which began in 2016 and demonstrated a working prototype last December, paving the way for a mass-produced version.

U.S. develops remote face recognition system to achieve target recognition within 1 km

It is reported that the current research is under way, the United States Special Operations Command did not say when.

The technology was designed for handheld devices and drones, and targets can even be identified without knowing the camera. As a result, human rights advocates have expressed concern.

The device is understood to have been developed by Secure Planet, based in Arlington, Va., which manufactures remote face recognition devices based on dslr sLR cameras and runs commercial face recognition software on standard laptops.

The identification range of these devices is typically about 300 meters, and extending the device’s recognition distance is not as simple as adding a longer lens to the camera, as it increases the noise generated by lens vibration.

The challenge for the new ultra-long distance recognition system is how to convert captured images into images that are clear enough to be used in passport processing software.

Researchers have tried to convert a series of blurry images into a single clear image using convolutional neural networks. Neural networks can also help extract pictures from complex environments.

Especially in the wild, the face may be at an inappropriate angle, moving, light poor, or being obscured by shadows, headscarves, sunglasses, or other obstacles.

Using neural network to identify faces, it is essentially based on the method of face feature extraction, the main features are extracted by multi-layer network convolution dimensions, and the input of sample pictures is used to automatically form a feature extractor and classifier, thus forming a model suitable for detecting and identifying targets.

Experts say this could be a step toward smupted weapons. In the future, such weapons could locate and attack targets without effective human control.