What is the failure of face authentication of China Everbright Bank?

Face authentication failed for several reasons:

1, lighting problem, whether the lighting intensity and uniformity at the time of registration and your current certification have changed greatly, and try to ensure similar lighting conditions as at the time of registration.

2, wearing glasses, that is, to ensure that it is the same state of wearing glasses when registering. I didn't wear it when I registered before, and I don't wear it now. If I've worn it before, I should wear it during the appraisal.

3, Liu Hai's occlusion, but also to ensure that it is consistent with the registration.

Face authentication refers to the biometric authentication of an individual through face recognition. In fact, the research on face recognition has a long history, starting from1960s. The recent attention to this technology is attributed to the use of artificial intelligence (AI) to improve the accuracy of authentication, thus greatly enhancing its functions.

The identification process is usually divided into two stages.

The first stage is to find the face from the image. Using artificial intelligence (AI), we first identify the areas that seem to be faces, and then the length and balance of each face (eyes, eyebrows, nose, mouth, etc.). ) is quantized to capture the characteristics of an individual for identification.

In the second stage, the pre-registered face image data is compared with the image data obtained by "face detection" earlier, and it is judged that the person is the person.

As a recent trend, 3D face recognition has been introduced. This is to use a 3D sensor to obtain the 3D information of the face, so surprisingly, it can be recognized from various angles (such as diagonal). In addition, it can also solve the darkness of illumination and facial expression, whether there are differences in makeup, beard and facial orientation, and most importantly, it can't cheat by using photos, so it can further improve the recognition accuracy and better ensure the safety of user information.

Today, with the widespread use of wireless Internet and smart phones, face recognition is also facing more and more challenges. In the unreliable collection environment, users' facial expressions are ever-changing, and their complexity far exceeds the pictures obtained in standard evaluation. Generally speaking, the difficulty of face authentication technology lies in:

1, the image resolution is low. In high-resolution images, face images can convey richer user information. In these Internet applications, the quality of cameras is mixed, so it is difficult to extract facial features effectively.

2. Modal diversity. Due to the difference of acquisition equipment and environmental light sources, there are many different ways of face images, such as ID card photos, digital photos, artistic photos and so on. Under different illumination bands, the reflectivity of human face will also cause the loss of image information.

3. Shooting environment such as lighting, posture, occlusion and expression. These factors have been difficult to solve in face recognition, such as yin and yang faces, black-rimmed glasses, grimaces and so on.

4. Inevitable external interference such as makeup and aging.