Face recognition will go into ordinary life and there are unsolved problems

Recently, Netease said it will adopt face recognition on its free email login security service. This makes a lot of network friendly, what role can face recognition play? What about practicality?

When a person walks to the front door of the vault, the camera scans his face. At this time, his facial features are the password to enter the vault. The door is automatically opened after verification. In the past, this was a Hollywood blockbuster. The scenes seen have now become a reality in life. Perhaps in the near future, faces will become important business cards and tags in the electronic life of ordinary people.

However, face recognition technology is not legendary. It also has its own limitations, and it still cannot achieve the 100% recognition rate expected by people.

■ Network “Preliminary” Face Recognition Today, the application of face recognition systems has been widely used and can be used in public security, airports, border defense ports, security and other important industries and fields, as well as smart access control, door locks, attendance and other civilian markets. This face recognition technology, dubbed the "Master of Solving Cases," "Home Security," "Attendance Helper," and other titles, has recently become more personal and is intended to be the first to meet the public on the Internet.

Recently, Netease said that it has achieved breakthroughs in the core technology of face recognition systems. It is expected to be the first to use Netease's free email login security application in the first half of this year. The user with the camera can independently choose to open and compare the face information captured by the camera.

In order to understand the relevant details, the reporter has repeatedly tried to get in touch with Netease, but as of press time, Netease has not received an interview. However, in the press release issued by Netease, we can see that in the actual use of face recognition system, NetEase mailbox users can choose to require face authentication when they log in at a different location, or face authentication after multiple passwords are locked in a short time. Unlock.

Relevant person in charge of NetEase also said that the recognition rate of NetEase's face recognition system is close to 100%, and it is almost impossible to misjudge, and it will not be deceived by photos. Even with changes in appearance caused by age, makeup, injuries, etc., the system can accurately identify differences in appearance under different environments and light.

â–  Wide range of face recognition applications In fact, face attendance, face access control, and face recognition are not simply photos taken by ordinary people and they are simple to compare. The contrast is behind a series of calculations and conversions. .

Professor Tan Xiaoyang of the School of Computer Science and Technology of Nanjing University of Aeronautics and Astronautics has long been engaged in the research of face recognition. He told reporters that from an application point of view, face recognition includes face recognition, that is, to identify the identity of a person based on face images. Whose problem is the solution? In addition, it also includes face identity verification and verification, that is, judging whether the face in the image is a designated person, that is, whether the problem is solved by someone.

The National 863 Face Recognition Project, Professor Su Guangda of Tsinghua University, introduced that after 40 years of development, face recognition has reached a certain level of maturity. Compared with other biometric identification technologies, face recognition has unique technical advantages in terms of practicality.

Su Guangda said that face recognition does not require biometric methods that the user cooperates with. Therefore, the operation is highly concealed and is particularly suitable for security, criminal monitoring, and criminal arrest. At the same time, it uses non-contact acquisition, is non-invasive and easily accepted. Fingerprint recognition often causes discomfort to people. Among them, the biggest advantage of face recognition is that it is more in line with humans' recognition habits and is somewhat similar to human-computer interaction. Therefore, people have high hopes for this technology and expect to meet the needs of various applications ranging from national public security, social security, financial security, and human-computer interaction.

■ Face recognition rate is affected by many factors. "The existing face recognition system can achieve satisfactory results under the conditions of user cooperation and acquisition conditions. However, if the user does not cooperate and the acquisition conditions are not ideal, In the next, the recognition rate of existing systems will suddenly drop. Face recognition technology is far from the practical level, and there are still many challenging issues that need to be resolved,” said Tan Xiaoyang.

For example, when comparing faces, there are differences between faces stored in the system. For example, shaving a beard, changing hairstyles, adding more glasses, and changing expressions may cause the comparison to fail. In other words, if a person makes a small change, the system may fail authentication. Today, it is conservatively estimated that the accuracy of face recognition technology can reach 95%, but it does not reach 100%. At the same time, for twins, face recognition is almost impossible due to too many similar features.

Zhang Xiaobo, a doctor of the Institute of Automation at the Chinese Academy of Sciences, also believes that lighting, posture, and decoration have an effect on machine recognition. At present, in the United States, the most advanced face recognition system in the world has a 1% false reading rate when testing.

â–  "Apply to appearances"

There are unsolved problems. Su Guangda added that nowadays, the occlusion problem is very serious for face image acquisition in non-matching situations. Especially in the monitoring environment, the monitored objects may wear glasses, hats, and other accessories, which may make the collected facial images incomplete, thereby affecting the subsequent feature extraction and recognition, and may even result in face detection algorithms. Failure. How to effectively remove the influence of occluders is a very urgent research topic.

At the same time, facial appearance changes with age, especially for adolescents. For different age groups, the recognition rate of the face recognition algorithm is also different. The influence of age changes on face recognition algorithms must also be resolved.

In addition, although the performance of face recognition algorithms will decline as the size of face databases increases, how to maintain or improve the recognition rate of face recognition algorithms in large-scale application environments is also a very important issue.

In fact, the advent of any technology has many bright spots that are worth looking forward to. However, there is always an inevitable gap between ideals and reality. All these need to take every step rigorously, and only gradually approach the ideal.

â–  Text / Li Xin â–  Tips

The face recognition process is first of all face detection, that is, judging whether there is a face in the input image, and if so, the position and size of each face are given. Followed by facial features positioning, that is, for each face found, the detection of the position and shape of its main organs and other information, and finally face comparison: According to the results of facial features positioning, compared with the face of the library to determine the face Identity information.

In terms of security, face recognition system information storage is still based on a computer-recognizable language, that is, numbers or specific codes. This means that it will also face hacking attacks. However, if the stored face information is encrypted, even if the hackers steal the "face", there is no way to identify or open it.