Face recognition rate has reached 95% recognition rate, smart home industry chain will be upgraded

This is not a fragment of the American drama "The Tracker of the Suspect", nor is it the future of a science fiction written. Face recognition technology has been used in people's daily lives. According to media reports, at the safe island of the Xinzhou Lotus intersection in Shenzhen, a huge display is particularly eye-catching. It does not broadcast advertisements or movies, but photos of pedestrians at the intersection. This is the face recognition technology. As early as April 2017, Shenzhen used a face recognition red light capture device at the intersection of Xinzhou Lotus. The device automatically used the camera to shoot pedestrian red light behavior and instantly played on the big screen on the safe island.

Algorithm + 2D, 3D face recognition technology

The status of face recognition was first developed in 1993. The US Department of Defense Advanced Research Projects Agency and the US Army Research Laboratory FERET project team established the FERET face database to evaluate the performance of face recognition algorithms.

From a broad perspective, face recognition includes the following five aspects: face location and detection, face representation (face feature extraction), face recognition, expression/attitude analysis, physiological classification, and so on. Face recognition has been rapidly developed in the field of biometrics due to its acceptability.

The convergence trend of big data and face recognition technology, in the era of big data and mobile internet, face recognition technology has been incorporated into Internet applications, especially mobile Internet applications. Large-scale face search based on big data is an important direction for the future development of face recognition technology. The large-scale face search technology developed at present can realize the rapid retrieval of billion-level face, the small memory required for a single face, and the high query efficiency. 360 image search, Century Jiayuan, etc. have begun to apply this technology.

The traditional face recognition method is based on two-dimensional image analysis theory for face analysis, feature extraction and recognition. Different from two-dimensional data, the three-dimensional data acquired by the face 3D scanner contains the spatial information of the face, which is the inherent information of the face itself. The shape data of the three-dimensional data does not change with the change of illumination and view, and the appendages such as makeup has a great influence on the image and has no obvious influence on the three-dimensional data.

The biggest shortcoming of the two-dimensional face recognition method is that it is vulnerable to face posture, different lighting conditions, facial expression changes and facial makeup. The accuracy of recognition is greatly limited, and these are the faces of the face at any time in nature expressed. 3D face recognition can greatly improve the recognition accuracy. Real 3D face recognition is researched using depth images.

The combination of two-dimensional and three-dimensional face recognition, the recognition of multiple modes can effectively improve the accuracy of face recognition; the two-dimensional recognition algorithm is gradually applied to three-dimensional face recognition; the face recognition algorithm can overcome: posture, expression Changes, wearing eyes, jewelry and other factors and light effects; recognition algorithms should require less computation.

Face recognition enterprise inventory

Domestic face recognition companies mostly use universities as the background to conduct technical research in conjunction with major research institutes around the world. Cloud from Technology, Face++ and Feisou Technology are currently leading the industry.

Backed by the "father of computer vision", the Chinese Academy of Sciences laboratory entrepreneurial team founded Yun Cong Technology. In addition to the alumni from the Chinese University of Science and Technology, the cloud from the science and technology team members also come from the world's top universities and research institutes such as the Chinese Academy of Sciences, UIUC, IBM, NEC, MicroSoft, etc.; and commercialization in the fields of finance, security, education, etc. Exploring, the IBIS platform has made it the largest supplier in the banking industry. The application of dynamic face recognition system in Guangdong has become a benchmark and has been promoted nationwide.

The Tsinghua Entrepreneurship Team launched the face cloud recognition open platform Face++. In addition to several Alumni from the Tsinghua University, the Face++ team also has research and development staff from Columbia University, Oxford University and the University of Southern California. And in the financial, security, retail areas, respectively, began commercial exploration and successfully developed vertical face verification solutions such as Face++Financial, Face++Security, Face++BI. Magic Man Camera is the world's first mobile phone application that makes real people into humorous comics. With more than 200 million users, Face++ automatically captures face and key detection technology and combines expression migration and image fusion technology to generate user's private comic image to create a variety of user styles.

Feisou Technology Co., Ltd. is a high-tech company that applies technological innovation, independent research and development, and applies machine learning, especially deep learning research results to face recognition, image recognition, and video content recognition. The real-time face recognition algorithm of Feisou is more accurate than 99.0% on the world public test benchmark dataset Labeled Faces in the Wild (LFW), which is more accurate than the 97.35% rate published by Facebook on the same face dataset in June 2014.

In addition to domestic companies, foreign Emotient, Affectiva and Amscreen are widely used in advertising, games and other fields.

Face recognition in the field of home use

The application of biometric technology in the field of smart home is becoming more and more popular. From the family as the embedding point, the intelligent home has opened the intelligent mode from the moment of entering the door. The face recognition door lock has emerged as the times require, and it has been produced in practical applications. Very good value. The so-called face lock is a door lock that identifies the person according to the facial feature information of the person. The user needs to register to be used normally, and the unregistered user cannot recognize the door opening.

In the development of smart homes, more biometrics will be applied to biometrics. Only by continuous technological innovation can the industry continue to develop and the value of biometrics will continue to rise in the future.

The construction of smart homes in China has entered a large-scale pilot stage. In the next 10 to 20 years, with the acceleration of the process of family intelligence in China, problems such as home and home are gradually emerging, and face recognition technology is the key technology to achieve a safe and secure home. Therefore, with the large-scale construction of smart cities, the application of face recognition technology will be a new trend in the future.

Summary

At present, the best face recognition system has been able to achieve a recognition rate of more than 95% in an environment with ideal environment and user cooperation. The application of the face recognition system is no longer far away from our lives. However, face recognition technology is still not a very mature recognition technology, but has a certain application within a limited scope.

The page contains the contents of the machine translation.