搜索
Search
新闻中心
/
/
/
Face recognition has entered the 2.0 era, pursuing more natural AI interaction
Check category

Face recognition has entered the 2.0 era, pursuing more natural AI interaction

  • Categories:Industry News
  • Author:
  • Origin:
  • Time of issue:2017-03-23 08:00
  • Views:

Face recognition has entered the 2.0 era, pursuing more natural AI interaction

(Summary description)With the further development of technology, iris recognition has a dedicated infrared camera and dedicated infrared light, which makes the recognition more feasible. In addition to identifying scan ray upgrades, machine learning and the iteration of the algorithm itself are also important aspects to solve the reliability problem.

  • Categories:Industry News
  • Author:
  • Origin:
  • Time of issue:2017-03-23 08:00
  • Views:
Information

With the further development of technology, iris recognition has a dedicated infrared camera and dedicated infrared light, which makes the recognition more feasible. In addition to identifying scan ray upgrades, machine learning and the iteration of the algorithm itself are also important aspects to solve the reliability problem. Lin Yuanqing said that data, algorithms, and products are a positive cycle. The iteration of data and the evolution of algorithms promote the production of better products, and better products are counterproductive to the further improvement of data and algorithms. "In fact, the breakthrough of Baidu's face recognition technology is inseparable from the massive investment in algorithms in recent years." Lin Yuanqing said, "Baidu entered the face recognition industry with 

1

"Face brushing" and "Rainbow consciousness" are loudly called

Li Ziqing, director of the Research Center for Biometrics and Security Technology of the Institute of Automation, Chinese Academy of Sciences, told the author that non-contact biometric technology refers to the technology that can collect identity information without direct contact with the user, such as iris recognition, face recognition, and voiceprint. Recognition, retina recognition, etc. belong to this type. Unlike voiceprint recognition that requires higher sound environment and the inconvenience that retinal recognition may bring to people with vision correction, the industry generally believes that the application scenarios of face recognition and iris recognition are more extensive and convenient, and its commercialization prospects are more worthwhile expect.

  Iris recognition has come into everyone's daily life. At present, more than 50 airports around the world have installed systems that allow passengers to pass customs directly through iris recognition. Singapore has also adopted "iris scanning" this year as another identification basis for travelers and Singapore citizens in addition to photos and fingerprints. Iris recognition is also showing its skills in the financial field: In order to prevent the traditional bank’s "username + password" information database from being stolen by criminals, some banks have begun to add iris recognition verification, that is, only when the iris scan information and the account information completely match, the customer will You can obtain the authorization for the withdrawal of funds, and the security level is fully upgraded.

   Compared with fingerprints and digital passwords, iris recognition is more secure in Internet payment applications, which also makes the iris scanning system gain the favor of Samsung and other brands of mobile phones. According to Qiu Xianchao, deputy general manager of Beijing Zhongke Hongba Technology Co., Ltd., the "iris phone" may become an indispensable "smart organ" for people in the future.

However, the public expects more face recognition than iris recognition: "Sweeping face to eat" and "Sweeping face pass" can really be realized? "Now in Baidu Science and Technology Park, employees can pass the "Sweeping face" in 1 second. "Passing gates', no work card, no need to wait." Lin Yuanqing, director of Baidu's deep laboratory, described the application of Baidu's face recognition technology to the author: "The next application scenario is likely to be Baidu canteen-Baidu employees can go to the canteen to eat. Say goodbye to work cards, and tens of thousands of people will be the first to enjoy the convenience of'brushing their faces'."

   Pursue more natural AI interaction

   There are two different scenarios for face recognition, one is 1:1 face comparison, and the other is 1:N face recognition. The former is commonly used in some applications of banks and public security systems. It is generally necessary to submit personal ID card information and facial features, and then the system will compare the ID photo and the submitted information with a 1:1 comparison; the latter has also been used in criminal investigation and other fields. Certain applications, but due to the 1:N identification accuracy problem, the difficulty is high. Traditionally, the public security system only uses this technology to assist in the investigation.

   Baidu's "face brushing" technology chooses 1:N precise recognition. Lin Yuanqing revealed that Baidu began to work on 1:N recognition applications in 2016 and has made great technological breakthroughs. "It has been possible to achieve very accurate 1:N face recognition." Baidu previously announced that the accuracy rate of Baidu Brain's face recognition and monitoring has reached 99.7%.

"If the accuracy rate is not enough, Baidu employees may be blocked from the gate in the application of face-swiping and passing gates. This situation is intolerable." Lin Yuanqing told the author that Baidu's face recognition has two accuracy. Two indicators: false pass rate and pass rate. "The pass rate is easy to understand and refers to the correct rate of matching. The false pass rate refers to'people who are not Baidu let it go easily'. Our index is close to 0; the pass rate is above 99%."

   In fact, 1:N accurate recognition is not easy to achieve high-precision recognition. This also makes scholars such as Li Ziqing doubt: "Is it really as high as the propaganda?" "1:N is the most difficult to achieve high-accuracy recognition." Lin Yuanqing also admitted, especially with The increase of the database-the current Baidu pass gates and the future Baidu canteen's database will be 20,000 to 50,000 people-poses more severe challenges to the accuracy and safety of recognition.

"But this is Baidu's "dish"." Lin Yuanqing said that only by achieving 1:N accurate recognition can it truly bring revolutionary changes to the original face recognition methods and make the interaction in the era of artificial intelligence (AI) smarter. , More convenient. "We must do this, because we hope that the future AI interface will be very natural and seamlessly connect with our lives."

   quietly entered the 2.0 era

   In fact, when face recognition first moved towards application, the "accuracy rate" has always been the bottleneck of deepening the application. In particular, the original two-dimensional face recognition cannot achieve a high degree of matching between "face" and "person". "Face tracking and recognition technology has entered the 2.0 era." Ye Zhou, the founder of artificial intelligence technology company ULSee, told the author: "Traditional face recognition technology is mainly based on visible light images, but this method has insurmountable defects. , Especially when the ambient light changes, the recognition effect will drop sharply, and there are other problems such as the large difference between the side face and the front face, and the use of photo images for fraud."

A product manager of a company in the field of biometrics told the author, “Nowadays, live body detection replaces static image recognition. Infrared light is used as a means of collecting facial features to make up for the shortcomings of visible light recognition. Real masks, photos, and mobile phone videos are not available. 'Muddy pass.'" Moreover, the product manager said that today people face face scanners and no longer need to "shake their heads" and "interact" with machines to ensure the authenticity of the input information.

   At present, the face key point detection technology can accurately locate the key areas of the face, and can even support a certain degree of occlusion and multi-angle faces, which greatly improves the accuracy. Even so, twin recognition and recognition before and after plastic surgery are still the technical bottleneck of face recognition. In this regard, Li Ziqing told the author that iris recognition may become one of the breakthroughs to solve this problem.

Different from the uncertainty displayed by the face shape at different stages of human life, the iris has a high degree of certainty and uniqueness. From fetal development to physical death, it always maintains its shape certainty and other individuals throughout the process. The difference. This property of iris has long been valued by biometrics experts. As an important part of non-contact biometric recognition, iris recognition also does not require direct contact between the eyes and the device, and is more convenient and faster. The dual authentication of the iris and the face is "double insurance" for face recognition.

   However, the average human iris diameter is 12mm, and the average Chinese iris diameter is even smaller. The small exposed area makes it difficult to collect iris feature information. Li Ziqing said, “In addition, the richness of the iris color of Chinese people is far lower than that of Westerners, which is also one of the problems faced by iris recognition.”

   With the further development of technology, iris recognition has a dedicated infrared camera and dedicated infrared light, which makes the recognition more feasible. In addition to identifying scan ray upgrades, machine learning and the iteration of the algorithm itself are also important aspects to solve the reliability problem. Lin Yuanqing said that data, algorithms, and products are a positive cycle. The iteration of data and the evolution of algorithms promote the production of better products, and better products are counterproductive to the further improvement of data and algorithms. "In fact, the breakthrough of Baidu's face recognition technology is inseparable from the massive investment in algorithms in recent years." Lin Yuanqing said, "Baidu entered the face recognition industry, taking a technical route."

1

Keyword:

Welbell | Face recognition temperature measurement terminal, science and technology help prevention and control

Welbell | Face recognition temperature measurement terminal, science and technology help prevention and control

"Face Recognition" Smart Cloud Access Control--Welbell Helps the Construction of Smart Community

"Face Recognition" Smart Cloud Access Control--Welbell Helps the Construction of Smart Community

Welbell attended the 2019 China Security Expo with many smart products

Welbell attended the 2019 China Security Expo with many smart products

这是描述信息

Telephone:0755-86006635
E-mail:sales@welbell.com

Address: High-tech South Seven Road, Science and Technology Park, Nanshan District, Shenzhen Area B, 3rd Floor, Building B2, Technology Park

Copyright: Shenzhen Welbell Intelligent Technology Co., Ltd.    粤ICP备09032168号  Powered by www.300.cn

Shenzhen Laiji Electromechanical Equipment Co., Ltd.