The dynamic living body detection method has high security, but requires the user to cooperate with the specified action, and the actual user experience is poor. In order to achieve the effect of non-sensing passage, face recognition access control rarely uses motion detection in response to instructions, and living body detection is usually carried out based on the difference between the image and the optical effect.

1. Ordinary camera live detection

Although there is no action response to commands, the real face is not static. There are always some micro expressions, such as the rhythm of the eyelids and eyeballs, blinking, the expansion and contraction of the lips and cheeks around them. At the same time, the reflection characteristics of real human faces are different from those of attack media such as paper, screens, and three-dimensional masks, so the imaging is also different. Univision cooperates with detection based on features such as moiré, reflection, reflection, and texture, and the detection system can easily respond to attacks from photos, videos, and prosthetics.

Using a specific physical feature, or the fusion of multiple physical features, we can train a neural network classifier through deep learning to distinguish whether it is a living body or an attack. The physical characteristics in living body detection are mainly divided into texture characteristics, color characteristics, spectral characteristics, motion characteristics, image quality characteristics and heartbeat characteristics.

Texture features include many, but the most mainstream ones are LBP, HOG, LPQ, etc.

In addition to the color characteristics of RGB, academic circles have found that HSV or YCbCr has better performance in distinguishing between living and non-living bodies, and is widely used for different texture features.

The principle of spectral characteristics is that biological and non-living entities have different responses in certain frequency bands.

Motion feature extraction target change at different times is an effective method, but it usually takes a long time and cannot meet real-time requirements.

There are many ways to describe image quality characteristics, such as reflection, scattering, edge, or shape.

2. Infrared camera live detection

Infrared face detection is mainly based on optical flow method. The optical flow method uses the time domain change and correlation of the pixel intensity data in the image sequence to determine the "motion" of each pixel position, that is, obtains the operation information of each pixel from the image sequence, and uses the Gaussian difference filter and LBP features. The support vector machine performs statistical analysis on the data.

At the same time, the optical flow field is more sensitive to the movement of objects, and the optical flow field can be used to evenly detect eye movement and blinking. This living body detection method can realize blind detection without user cooperation.

From the comparison of the above two pictures, it can be seen that the optical flow features of the living face are irregular vector features, while the optical flow features of the photo face are regular Orderly vector features can distinguish between living bodies and photos.

3. 3D camera live detectionThe face is captured by the 3D camera to get the 3D corresponding to the face area data. According to these data, select the most easily distinguishable features to train the neural network classifier, and finally use the trained classifier to distinguish between living and non-living. The selection of features is very important. The features we choose contain global and local information. This choice is conducive to the stability and robustness of the algorithm.