info@elefinetech.com
Tel:+8613534079085
Before we know it, face recognition has entered our lives. Not long ago, all the dormitories of Beijing Normal University installed a face recognition access control system, no matter who enters the door, you must swipe your face to let you pass; a restaurant in Hangzhou also launched a face recognition payment, the whole process does not exceed 10 seconds…So, face Where can recognition be applied, what is its principle, and what is its true effect?
Face recognition can first be applied to security monitoring. With the advancement of “safe city construction”, more and more high-definition cameras are deployed in various public places, such as airports, subways, train stations, and bus stations. Facial recognition software can be installed in these cameras to automatically detect The face in the video screen is compared with the face data in the database one by one to find and hunt criminal escapers.
In addition, face recognition can also be applied to a wide range of fields from access control, device login to individual recognition. For individual identification, face recognition can be used to verify ID cards, driving licenses, passports, visas, ballots, etc.; it can be used to control device access, vehicle access, smart ATM, computer access, program access, network access, etc. ; It can also be used for smart card user verification, face database face retrieval, face tagging, face classification, multimedia management face search, face video segmentation and splicing, and man-machine interactive games, active computing, etc.
These situations are not only the reality that face recognition technology will or have been widely used in people’s lives and work, but also a sign of the rapid development of bioinformatics technology. In the final analysis, face recognition is just one type of biometric identification technology. It is still too early to judge whether it can fully replace other biometric technologies, such as fingerprint recognition and iris recognition, and needs to be tested in practice.
However, the current face recognition technology is mainly used in three categories. The first is 1:1 authentication, where the certifier and the certificate information are unified, and it is mainly used for real-name verification. The second is 1:N certification, that is, to determine whether a person is a member of a specific group, which is used for personnel access management and urban security. The third is living body testing to ensure that it is a real person operating the business, and then to do the account license authorization. Obviously, face biometrics related to public life are mainly the first and third types, and the second type is mainly used for criminal forensics and anti-terrorism.
Face recognition is popular and popularized because of its huge advantages. Face recognition has biological natural attributes and simplicity. Of course, there are voice recognition and body shape recognition for the recognition of biological natural attributes, but it is not as simple as looking at the face recognition. The simplicity of face recognition is also reflected in non-contact and non-compulsory, which can obtain the recognized face image information without being noticed by the individual. Face recognition is the use of visible light to obtain facial image information. Unlike fingerprint recognition or iris recognition, these collection methods require the use of electronic pressure sensors, scanners, etc., which are easy to be detected by people and may be disguised and deceived.
In addition, face recognition is more efficient, and multiple faces can be sorted, judged, and recognized simultaneously in practical application scenarios.
Nevertheless, face recognition also has blind spots and weaknesses. The Mi 8 mobile phone uses face recognition and unlocking technology. However, after experiments, someone found that the face recognition of the Mi 8 mobile phone was easy: as long as there is an infrared photo of the owner, and the photo is not reflective, the phone can be unlocked. Moreover, print out the ordinary color photo of the phone owner with a black-and-white printer, and use a pencil to blacken the eyes of the photo and the shadow on the face to unlock the phone.
Therefore, although face recognition has the advantages of high efficiency, speed and non-invasiveness, the shortcomings of face recognition are also obvious. Moreover, for twins, before and after plastic surgery, sudden thinning and fattening of people, changes in appearance caused by aging, etc., face recognition technology may not be effective, or it may not be recognized, or it may cause misrecognition. More importantly, criminals and criminals can easily forge to deceive facial recognition technology. The face recognition system mainly includes four components, which are face image acquisition and detection, face image preprocessing, face image feature extraction, and matching and recognition. The current situation is that these four parts may have loopholes. As long as a person provides photos, even the simplest selfies, or criminals stealing photos of the target, 3D modeling can be used to change the static of the photo to dynamic with the help of key point positioning of the face and automatic face dynamics technology. The main actions are blinking, smiling, raising eyebrows, turning the head left and right, etc., which eventually deceive the face recognition system and bring great harm to people’s safety.
The vulnerabilities of the face recognition system have long been concerned by the computer field. There are four major conferences in the field of international computer information security, namely Oakland, CCS, USENIX, and NDSS. At the 2016 USENIX Security Annual Meeting, relevant professionals demonstrated that they only need to reassemble the photos collected on social media into a three-dimensional virtual avatar of a person, and then use the VR display to make it really come alive. Fooled the face recognition system.
However, the vulnerabilities of the face recognition system do not stop there. Just as there are no two identical green leaves in the world, it is impossible to have two completely similar human faces in the world. Even the faces of identical twins cannot be exactly the same. However, we have to admit that both in physical and In the biological world, things always have similarities, as do human faces.
The structure of all human faces is generally composed of the upper and lower jaw, cheekbones and various muscles, such as facial muscles (expression muscles) and frontal muscles. They have certain similarities, especially for the same race. The similarity is relatively large, which poses a challenge to the accuracy of individual identification.
The shape of the human face is very unstable, and people can produce many expressions through changes in the face. The visual images of the human face are also very different at different viewing angles. Moreover, lighting conditions (such as day and night, indoor and outdoor, etc.), coverings (such as masks, sunglasses, hair, beards, etc.), age, etc., can all affect the accuracy of face recognition.
Similarity and instability are both dynamic changes. The former is called inter-class change, and the latter is called intra-class change. The intra-class variation of a face is often greater than the inter-class variation, and the combination of the two makes individual identification more difficult.
Now, after face recognition vulnerabilities frequently appear, professionals have put forward another biggest danger of more individual biometric technologies, including face recognition, that is, the theft of individual biometrics. Like biometrics such as fingerprints and iris, the face is also a unique biometric feature. All biometric data entering the computer will be converted into numbers of 0 and 1 and stored in the database. The probability that these biometric data, which are regarded as unique, is stolen after entering the network is greatly increased, and the risk is much more serious than theft. This is the biggest insecurity of biometric authentication.
Because the biometric identification technology represented by face recognition is not mature at this stage, it is very easy to be stolen, and it is not suitable for widespread use on the Internet. However, it can be used in local or regional areas without the Internet, such as access control , Safes and bank vaults.
Moreover, using it in a local area can also expand the scope of face recognition, such as individual recognition in the college entrance examination, to prevent cheating. Because face recognition is a high-tech method that combines biology and computer technology, it mainly includes face image acquisition and detection, face image preprocessing, face image feature extraction, and matching and recognition processes. The first few parts of these processes can be carried out for each candidate before the college entrance examination, and then when the college entrance examination enters the venue, they can enter the venue only by face recognition, and based on the exam room number and location, supplemented by an admission ticket , ID card seated for examination.
The biggest advantage of this approach is that the installation of a face recognition system in the college entrance examination examination room for individual identification has great reliability, which can not only avoid the loss of identity documents and delay students to participate in the college entrance examination, but also efficiently identify the identity of the candidates, which can be avoided. Take the exam and take the exam to ensure the fairness and justice of the college entrance examination.
Face recognition needs to improve and adopt multiple recognition methods to ensure safety. For example, when it comes to privacy, property, finance, and payment circulation, a combination of facial recognition and voiceprint, fingerprint, iris recognition and other biometric authentication signals can be used. One more procedure is one more security.
In the future, it is necessary to carry out technological transformation and upgrading of face recognition technology to ensure the uniqueness, reliability, security and non-appropriability of individual identification.
However, even the best technology faces a problem, it needs to be operated by humans, and people’s lives cannot be completely controlled by technology, including biometric identification and safe operation. It requires human participation and integration with technology at all times. Just like the verification by the customs, the customs declarer will both “swipe his face” on each entry and exit, but also look at the person. Because the visual cortex area of the brain and the fusiform face area (FFA) are more effective than all biometric recognition technologies, and it only takes a second or two to determine the authenticity of the human face, whether it is the same person or a fake person.
Download Brochure