News

Near infrared face recognition

Face recognition is the most commonly used technology in the field of biometrics, which has been widely used in the field of public security in recent years. The key of all kinds of face recognition methods is to extract identity related features from face data, and eliminate the part affected by non identity factors, which generally include: environmental lighting, posture, expression, ornaments and so on. Among them, illumination is the most important problem in practical application. Users generally require face recognition system to adapt to different lighting environment. The general face recognition system uses the common visible light face image for recognition, this kind of system is easy to receive the influence of environmental light changes, before recognition, it often needs to use some preprocessing algorithm to process the light. Although the illumination preprocessing algorithm can eliminate the influence of illumination to a certain extent, it will also make the image lose some useful information.

Near infrared face recognition is a solution to solve the problem of illumination in face recognition. It includes two parts: active near infrared face imaging device, active near infrared face imaging device and active near infrared face imaging device; And the corresponding illumination independent face recognition algorithm. Using the active near-infrared light source with higher intensity than the ambient light and the optical filter of the corresponding band, the face image independent of the environment can be obtained. The face image will only change monotonously with the distance between the person and the camera. Some special feature extraction methods, such as local binary pattern (LBP) feature, can be used to further eliminate the monotonic change of the image and obtain the completely illumination independent feature expression.

In order to weaken the influence of ambient light on imaging, the intensity of active light source needs to be higher than that of ambient light, but strong light will interfere with human eyes and reduce user comfort. Therefore, active light source is best located in invisible band, such as infrared, ultraviolet and so on.

Long term ultraviolet radiation is easy to cause permanent damage to people’s skin and eyes. Mid and far infrared band imaging will lose most of the information of the object’s surface, which is generally not used for object imaging. Therefore, near infrared band becomes the best choice, such as 780nm.

Active near infrared face imaging

The purpose of developing active near infrared face imaging equipment is to provide face recognition with high-quality face image which is not affected by ambient light. The so-called high-quality image includes: appropriate brightness, uniformity, appropriate contrast, no over exposure, etc.

Active near infrared face imaging equipment generally includes the following parts:

1. When the intensity of the corresponding band is higher than the ambient light, the active near infrared light source and the camera are in the same position;

2. Camera capable of receiving near infrared light;

3. Narrow band filter, placed outside the camera lens, allows near-infrared light to pass through while filtering ambient light.

As the active light source will attenuate with the increase of distance, the distance between people and the light source (camera) is generally 50-100 cm.

The influence of illumination on near infrared face recognition

When the distance between the human and the light source is constant, the near infrared face image is very stable, but the overall brightness of the image will still change monotonously with the change of the distance, so it is necessary to use specific feature extraction methods to solve the monotonic change problem, such as histogram equalization, square Icon setting and so on. Local binary pattern (LBP) is a better scheme. LBP only focuses on the size relationship between pixels. When the whole image changes monotonically, the size relationship between pixels is unchanged.

After obtaining the illumination invariant feature expression through the above methods, the illumination independent face recognition will come naturally. Then, a set of illumination invariant near infrared face recognition system can be constructed by using the general feature selection and classifier design.

The disadvantages of near infrared camera for face recognition are as follows

Although near infrared face recognition has obvious advantages over traditional visible face recognition, the use of active light source also brings some problems

1. The active light source will produce obvious reflection on the glasses, reducing the accuracy of eye positioning;

2. The reflection can block the eyes and affect the recognition;

3. Users need to rebuild the near-infrared photo library instead of using a large number of visible photos;

4. After long-term use, the active light source will be damaged and attenuated, which will bring more later maintenance.

Future research will be devoted to solve the above problems, among which problem 3 is the most important, which can greatly expand the application scope of near infrared face recognition technology. It has gradually developed into an independent research direction “heterogeneous face recognition”.