News

Principle of Anti-facial Recognition Privacy Glasses

Strictly speaking, anti-facial recognition privacy glasses should be called anti-face detection collection glasses.Different from other algorithmic disturbance solutions, this one disturbs the camera’s correct collection of facial information from facial collection and detection in the whole work process of facial recognition system. The collected facial information fails or is short of much key recognition information. As a result, the subsequent facial recognition rate declines or fails, thus effectively preventing the normal facial recognition of the system.
At present, there are three kinds of facial recognition collection technologies: visible light recognition technology, near infrared recognition technology and 3D imaging recognition technology.

1）Visible Light Recognition Technology

The visible light facial recognition technology is a biological recognition + smart video analysis technology. The facial biological characteristics are collected through video. In any visible light and dark environments, the analysis comparison is done with the database data. The legal and blacklist users are accurately recognized. The accurate recognition is realized in the outdoor visible light environment for high-end users, including prison, army, government, police and electric power. The original backward biological technology recognition and passive monitoring security pattern in the world are completely broken through. The completely active prevention at active high-security level is created.
The greatest advantage of visible light recognition technology lies in its combination with the existing video monitoring system. It can carry out the personnel management for the long-distance and large-scope specific places in an intelligent, efficient, accurate and active manner. Besides, it can carry out the image collection of target crowds unconsciously, so it has stronger concealment. However, its disadvantage lies in the great reduction of recognition precision or even failure of recognition at night or when the light is poor or the target personnel has any shelter.

2）Near Infrared Recognition Technology

The near infrared facial recognition is a solution to the illumination problem in the facial recognition. It includes two parts: active near infrared facial imaging device and corresponding illumination unrelated facial recognition algorithm. Through the active near infrared light source imaging whose intensity is higher than ambient light, and the optical filter disc of corresponding wave band, the environment unrelated facial image can be obtained. The facial image will have the monotonic change with the change of human and camera distance. In such image, some specific characteristic extraction ways are adopted, for example, LBP pattern, the monotonic change of the image can be further removed, and the completely illumination unrelated characteristic expression is obtained.
The greatest advantage of near infrared recognition technology lies in the elimination of impacts of light on recognition effect. However, its disadvantages are also conspicuous: 1. In order to reduce the impacts of ambient light on recognition effect, the intensity of infrared active light source should be higher than the ambient light intensity, which greatly reduces the recognition distance. 2. The active light source will generate an obvious reflection on the glasses, which reduces the eye positioning precision. 3. The existing visible light pictures cannot be made use of, and the user has to re-construct the near infrared picture library, which consumes much time and energy. 4. After the long-term use of active light source, damage and attenuation will occur, which causes more maintenance in the later stage.

3）3D Imaging Recognition Technology

3D imaging recognition technology is an important development and discovery of facial recognition. Presently, most facial recognition applications are limited to 2D image. In essence, face is a 3D model. 2D facial recognition is easily impacted by posture, illumination, expression and other factors, for 2D image has a defect, that is, it cannot express the in-depth information favorably. If the in-depth learning is to understand the facial recognition from the cognition angle of human, 3D technology reflects the facial recognition from the realistic model.
Currently, the arithmetic research on 3D facial recognition is not so rich and deep as 2D facial recognition technology. Many factors have restricted the development of such technology. At first, 3D facial recognition often needs a specific collection device, for example 3D camera or binocular camera. Such collection device is expensive presently and it is mainly used for the specific scenario. Secondly, 3D modelingprocess needs a large amount of computation and has tough hardware requirements, which restricts the current application. Thirdly, 3D facial recognition database is scarce. The researchers lack training samples and test samples and cannot conduct the deeper theoretical research.
Currently, the common 3D imaging modules include: 3D structure light recognition module, 3D TOF recognition module and binocular 3D imaging recognition module.
Compared with the traditional camera, 3D camera has introduced vcsel module on the front end regarding the greatest difference in the hardware. 3D camera has the following features: it can obtain the plane image and the in-depth information of shooting object, namely, 3D position and size information. It is composed of multiple cameras + in-depth sensor generally.
The structure light scheme is suitable for front 3D imaging of consumer electronic product, and it is used for the near distance scenario. However, ToF scheme is suitable for rear 3D imaging of consumer electronic product, and it is used for the long distance and outdoor strong disturbance environment. As the detection scope is too small (less than 1m) and there are many long distance detection problems, the binocular 3D visual scheme has few application scenarios.
Our anti-facial recognition privacy glasses adopt the unique optical means and successfully intercept the transmission of some visible light, infrared light and infrared ray laser, and disturb the camera’s collection and detection of eyes and surrounding key facial information. Therefore, the recognition rate of facial detection system declines or even fails. Besides, the database’s correct record of new facial information can be disturbed.