In the past two years, with the rapid development of science and technology, face recognition has gradually become an important field of application of biometric technology in the new era. Forget the password? It's all right. We can also "brush our faces"! Today, Xiaobian will take you to know the latest face recognition technology, to see where this technology has developed.

Traditional face recognition technology is mainly based on visible image, and people are familiar with such recognition methods. However, the shortcomings of this method are very obvious. The limitation of light is very large, and it can not meet the actual needs. The solution to illumination problem includes three-dimensional image face recognition and thermal imaging face recognition. But these two technologies are far from mature and the recognition effect is not satisfactory.

A rapidly developing solution is multi-source face recognition technology based on active near infrared image. It can overcome the influence of light changes, and has achieved excellent recognition performance. The overall system performance in accuracy, stability and speed is better than that of three-dimensional image face recognition. This technology has developed rapidly in the past two or three years, making the face recognition technology gradually become practical.

At present, there are three main schemes in the market, they are 3D Structured Light, ToF 3D (Time Of Flight) and Binocular Stereo System.

3D Structured Light

The basic principle of 3D structured light technology is that by near infrared laser, light with certain structural characteristics is projected onto the object to be photographed, and then collected by a special infrared camera. This kind of light with a certain structure will collect different image phase information according to different depth regions of the object, and then convert the change of this structure into depth information through the operation unit, so as to obtain three-dimensional structure. Simply put, the three-dimensional structure of the object is acquired by optical means, and then the acquired information is applied more deeply.

ToF (Time Of Flight)

ToF is one of the three-dimensional depth cameras. It's a disciple of structured light. There are two kinds of ToF ranging: single point and multi-point. In general, multi-point ranging is used on mobile phones. The principle of multi-point ranging is similar to that of pulse single-point ranging, but its light-receiving device is CCD, i.e. charge-retained photodiode array, which has integral characteristics of light response. The basic principle is that the laser source emits a certain field of view angle laser, in which the laser duration is DT (from T1 to t2). Two synchronous trigger switches S1 (t1 to t2) and S2 (t2 to T2 + dt) are used for each pixel of CCD to control the time period when the charge-holding element of each pixel collects the reflected light intensity, and the response C1 and C2 are obtained. The distance between the object and each pixel is L = 0.5 * c * DT * c2/(c1 + c2), where C is the speed of light (the formula can remove the influence of the difference of reflector reflectivity on ranging). Simply put, sending out a processed light will reflect back when it touches the object and capture the time back and forth. Because the speed of light and the wavelength of modulated light are known, the distance of the object can be calculated quickly and accurately.

Stereo System

Binocular stereo imaging (Stereo System) uses two cameras to take pictures of objects, and then calculates the distance of objects through the triangle principle. Huawei nova3, which has been listed, adopts binocular 3D face recognition scheme and is the standard provided by IFAA (Internet Financial Identity Authentication Alliance), which is also the main reason why it reaches payment level. Two RGB (color camera) cameras on nova3 identify device users by simulating binocular vision systems. The principle of binocular scheme is simplest. It should be an earlier face recognition scheme with the lowest cost. All the depth information collected depends on the image captured by the camera and is obtained by software algorithm. The accuracy requirement depends on the capture resolution. At the same time, because this scheme relies on the algorithm to analyze the image to get the depth information, the calculation load is the largest, the algorithm complexity is the highest, the implementation is difficult, and the recognition speed is slow. It is also affected by light, especially in dark environments.