The face and development trend of future face recognition in the era of “brushing face”

In September, the face-to-face news frequently entered the public view: the face recognition access control system was installed in all the dormitory buildings of Beijing Normal University; a KFC restaurant in Hangzhou could pay for the face, and the whole process was no more than 10 seconds...

Does the face recognition technology application “outbreak” mean that documents, traditional identification methods such as fingerprint recognition will be replaced? Some netizens are anxious to ask: Fat? How to do it? What should I do with plastic surgery? What if someone pretends to be himself?

In recent years, the application scene of face recognition has been gradually expanded from online use to offline. There are unmanned supermarkets in Beijing that can brush in and out, Henan has an airport to face the face, Xiamen has a hotel to wash your face...

In June of this year, China Southern Airlines launched the face recognition boarding system at Henan Nanyang Airport, and passengers can board their faces.

According to data from the Prospective Industry Research Institute, the market size of China's face recognition industry in 2016 has exceeded 1 billion yuan, and it is expected to reach 5.1 billion yuan by 2021.

Three application directions of face recognition technology

According to reports, face recognition technology has three main application directions. An application scenario is 1:1 authentication, that is, the certificate and the certificate information are unified, and are mainly used for real-name verification.

Followed by 1:N certification, which is to determine whether a person is a member of a specific group, for personnel access management and urban security. A public security bureau in Chongqing used the portrait comparison system of Shangtang Technology to identify 69 suspects within 40 working days, which is 200 times higher than the artificial efficiency.

The third is in-vivo testing to ensure that the real person is operating the business and thus doing account licensing. Cloud from technology's face technology has been applied to more than 50 banks across the country. CITIC Bank (601998, stock bar) ATM and mobile client can remotely authenticate, Haitong Securities can open accounts remotely, and the Drip Platform can check whether the driver is a registered driver.

Face recognition accuracy has leap

Since the 1960s, scientists have been involved in face recognition research, but the accuracy rate has not been high. In the past few years, scientists have applied deep learning algorithms to face recognition. In 2014, the algorithm developed by the research team of the Chinese University of Hong Kong, Tang Xiaoou, made machine recognition more accurate than humans.

According to Yan Shuicheng, deputy president of 360 Company, Dean of Artificial Intelligence Research, face recognition usually includes the following steps: Camera or professional equipment first collects pictures, face detection technology locates the face in the picture, and then repositions such as the corner of the eye? The nose point, the corner of the mouth, the feature points such as the contour of the face, and corrections such as ray compensation or occlusion elimination. The deep learning algorithm is used to extract the identity features, and the face features in the database are compared to identify the face identity.

"With the adoption of deep learning algorithms, the accuracy of face recognition has leapt significantly compared to five years ago," said Yan Shuicheng.

China's R&D level is basically in sync with the international level of development, but the plastic surgery is too large, and the twins are still difficult to identify.

National “Thousand Talents Program” expert? Gong Yihong, Distinguished Professor of Xi'an Jiaotong University School of Telecommunications, said that there are several recognized test sets in the field of face recognition. The top team names in the test results will be recorded online. Chinese researchers occupy half of the country. Accuracy is over 99%.

Despite this, Gong Yihong said that the results obtained in the test set will be greatly reduced in the real world. In reality, the crowd sample is larger, and different lighting, attitude, resolution and other conditions may cause difficulty in machine identification.

Yan Shuicheng said that twins? Before and after special circumstances, the machine can identify the machine, depending on the specific situation. For example, if the cosmetic volume is too large, it is possible that the machine cannot be identified. In addition, facial information changes with age. If it is not recognized by the machine, the user can simply solve the problem by updating the face photo.

Experts predict: the future identification trend is "one person and one"

Experts said that compared with fingerprint recognition, iris recognition and other biometric methods, the biggest advantage of face recognition is "non-contact", which can greatly improve the system response speed, while avoiding the hidden dangers caused by fingerprint machine contact. In addition, the "non-cooperating? non-invasive" feature means that data can be collected without the cooperation of the user, which is beneficial to the application in the security field.

"Face recognition will become the main means of identification, and it can be used on a large scale within two or three years." However, Gong Yihong also believes that a mainstream trend in future identification is the integration of people and certificates, that is, brushing faces and ID cards. Will continue to exist.

Xie Yinan said that for government agencies and banking institutions, there are many businesses based on the principle of “knowing” and “self-to-be”, which may also be replaced by remote processing. For example, ID card processing, bank account opening, social security and notary business, etc. If you can verify your identity on the mobile phone, it means that the ordinary people do not have to go to the scene to handle it.

Face recognition technology also faces some security vulnerabilities

At present, through some special means, such as using other people's photos, etc., it is possible to successfully "cheat" the face recognition system. In this regard, many companies have increased their technical investment in “living detection”, and in the scenario of high security requirements, many companies require dual verification.

In addition, the collection of faces also increases the risk of privacy breaches. Xie Yinan said that the solution is to “desensitize” the photos when they are collected, extract only the photo features, and not save the photos themselves. This requires not only the self-discipline of the collectors, but also the establishment of a unified standard for the entire industry under the guidance of the government, and jointly build an industrial dam that protects the privacy of users.

P02-M Series Push Wire Connectors

Super thin, room saving, economically efficient

Inter lock structure, can be combined to desired number of poles

Multi-pole push-in connector

Mono/dual press button for easy wiring and releasing

Compatible with earth tag and/or cable clamp

Fixed Foot connector Terminal,Minitype Wire Connector,super slim fast Connectors,high performance fast connectors

Jiangmen Krealux Electrical Appliances Co.,Ltd. , https://www.krealux-online.com

Posted on