When will the 40-year “brush face†technology road safety problem be solved?
September 12, 2018, the emergence of a new generation of iPhone, one of the biggest highlights of Apple's new product is the new identity authentication system - "FaceID", that is, facial recognition, "brush face" that is unlocked.
Whether such a simple unlocking method has security risks? Apple executives stressed that face recognition unlocking is very safe, the phone uses a 3D form to authenticate the user's face. Only when the phone is in a stable state and the user looks directly at the camera, can the lock be unlocked. The system does not Masked by a mask or photo. When the user shifts his gaze or closes his eyes, the screen is automatically locked. The user's face information data will be stored securely in the phone's local storage and will not be uploaded to the cloud.
Face recognition belongs to one of the biometric identification technologies. It refers to distinguishing each individual from the biological characteristics of the organism (generally referring to humans). The identification of biometric features such as face, fingerprint, palm print, iris, retina, speech, body shape, and personal habits (such as signatures) has corresponding identification technology support. These features are often seen as a convenient form of authentication because they are mostly inherent and unique.
Research on face recognition technology began in the 1960s. At first, this technology was not fully automatic. The program will calculate the distances and proportions between the features, and then compare the data obtained with the reference values ​​already in the database. However, the program operator needs to manually “lock†facial features such as eyes, ears, nose, and mouth. Experiments can continue.
In the 1970s, researchers created a program that can automatically recognize human faces. This program uses 21 specific facial features such as hair color and lip thickness to determine the face. In the late 1980s, a breakthrough study found that to accurately identify a normal face, less than 100 specific variables were needed.
With the rapid development of computer technology in the 1990s and the need for the United States military to start funding related research for anti-terrorism, face recognition technology has only been quickly implemented from theory to application. The FEERT project funded by the US Department of Defense organized three face recognition evaluations in 1994, 1995, and 1996. Several well-known face recognition algorithms all participated in the test. The three evaluations directly promoted the face recognition algorithm. improvement of.
After the "September 11" incident, in order to curb terrorist attacks, the United States has paid more attention to face recognition technology and has been more widely used in the field of security. In the following decade, face recognition technology was limited to the field of security, and there has been no substantial breakthrough. Although such products have also appeared on the civilian market, they are mostly concentrated on security systems such as access control, attendance and monitoring systems. The only exception is that in 2006, Nikon took the lead in applying face recognition technology to products. The new camera automatically searches for human faces and prioritizes focus when shooting.
Turning points appeared in the era of mobile Internet, giants took the lead. At the end of 2010, Facebook (FaceBook) took the lead in introducing the "TagSuggestions" function using face recognition technology in the United States, and in June 2011 this feature was extended to most countries outside the United States. Google also launched a similar feature "FindMyFace" on Google+ in December 2011, allowing users to automatically find friends in their friends' photo albums through facial recognition.
In the following two years, the two companies wanted to acquire a number of face recognition technology startups, such as Google’s “Pittsburgh Pattern Recognition†(PittPatt), acquired in 2011; Face.com, an Israeli face recognition technology company acquired by Facebook in 2012. Enables people to automatically identify photos in the photos without labels, helping users to sort photos.
In 2013, Finnish startup Uniqul launched a payment platform based on the face recognition system. When paying, you only need to face the camera on the POS screen and click "OK". Uniqul also applied for a patent on this technology and claimed to have "military grade algorithm" protection. This is the first time the world has applied face recognition to the payment sector.
In December 2013, Facebook created a deep-learning artificial intelligence laboratory in New York and invited YannLeCun, the in-depth learning originator, to join. Under his promotion, Facebook's DeepFace technology face recognition accuracy reached 97% in 2014. In 2015, YannLeCun publicly stated that even if there are no positive faces in the images, the programs they develop can also determine their identities from the user's hairstyle, posture and body type. In early 2014, Google acquired DeepMind, a deep-learning algorithm company, and Jetpac, an image analysis company, for $400 million, followed by FaceNet, a face recognition technology.
In China, Baidu also recruited Daniel Wu, a senior in the field of deep learning, to join the establishment of the Deep Learning Institute in May 2014. In 2015, Baidu released Face Excellent Products. In March 2015, Ma Yun demonstrated the SmiletoPay brushing technique of Ant Financial to German Merkel and Chinese Deputy Makai at the Hannover Consumer Electronics, Information and Communication Expo, and purchased the 1948 Hannover commemorative stamp from Taobao.com.
With the improvement of mobile device processing capabilities, face recognition technology has rapidly broken through the security field, flocking to everyday applications, and fermenting in financial systems, entertainment and other fields, creating tremendous commercial value.
At present, deep learning is gradually being applied to face recognition. Deep learning combines the two steps of feature extraction and classification and uses the characteristics of the neural network black box to calculate the most suitable feature extraction mode. The drawback of extracting impact recognition results in making the algorithm more suitable for different practical applications.
Although there are many kinds of face recognition technologies, the underlying algorithms are very similar, and the “diversity†and “precision†of recognizing pictures are important standards for measuring the level of technology. Only when a certain amount of training data is “fed†to the machine to enhance its ability of deep learning, can it be ensured that the face recognition technology achieves the desired effect in practical application scenarios. This means that in order to improve the accuracy of the algorithm, a large amount of data accumulation is essential.
Who stole our "face pattern"
Taking the example of an internal surveillance camera in the shopping mall to illustrate, it is easier to understand the principle of “brushing the faceâ€: a computer equipped with face recognition software will detect and identify video images in shopping malls. Once the system finds any suspicious "face", it will pay close attention to each "face" in that shot. When the system adjusts the face in the image to the proper size and orientation, it will further identify it and create a “face patternâ€.
Faceprinting is similar to the principle of fingerprinting, that is, a set of combined features that can distinguish human faces. The geometrical relationship such as the distance, area, and angle between facial features such as eyes, nose, and mouth varies from person to person. A large eye, small nose, and thin lips “face lines†are similar to small eyes, large noses, and thick lips. People's face is different.
Comparing a face print with a photo can verify the identity of a “known personâ€, such as the company’s identification of employees entering a specific area. Facial lines can also be compared with a large number of pictures in the database to identify an "unknown person." However, when the light changes, the face is covered by foreign objects, and the facial expression changes, the characteristics change greatly. A dark mirror can greatly confuse the facial recognition system.
In addition, the degree of cooperation of the person being identified with the recognition process is also one of the factors that determine the success or failure of facial recognition. Therefore, it is relatively simple to recognize people who consciously perform facial recognition. For example, when paying for a face, the customer needs to look directly at the camera in a suitable light for the convenience of software recognition.
In order to solve these technical difficulties and improve the accuracy of facial recognition, companies that research and develop or apply face recognition technologies are accumulating large amounts of data to improve the system's deep learning ability. For example, as a partner of Mito Xiuxiu, Face++ has identified more than one billion photos and has now established hundreds of millions of celebrity photo galleries. In order to research DeepFace technology, Facebook has already established a 4.4 million tagged face pool from 4030 individuals. Facebook hopes to use deep learning algorithms to investigate users' behaviors and habits on social networks and push information precisely.
Smart hardware and cameras collect our personal image data anytime and anywhere. Accumulating user data for such a long time will inevitably involve personal data and privacy protection issues.
A mobile application called FindFace allows users to find accounts on social software by using facial recognition technology, just one person’s photo. On the surface, this is a great way to contact friends and colleagues, but this process can be easily misused, and people may use it to expose others or cause harassment.
In 2014, a professor at Carnegie Mellon University in the United States found that after Google's image search for an image of an anonymous marriage website user, it would be easy for humans to have real information about these users. Disney had received consumer complaints. It used Trapwire, a face recognition system, to obtain consumer credit card information and push products they might be interested in.
In July of this year, Wal-Mart was revealed that it was developing a facial recognition system to identify whether customers were unhappy and applied for a patent. Recently, two researchers at Stanford University have developed a neural network algorithm that can be used to identify a person's sexual orientation through face recognition, and the accuracy of the algorithm's test results is as high as 91%. The results of this research and the possible discrimination issues it may cause are deeply disturbing to the LGBTQ community.
Not to mention the use of face recognition technology and police portable cameras, positioning software, and other technologies that assist in real-time tracking of machine learning. They are very good at fighting crime, but they also expose the privacy of our normal lives. Under the public power.
For a long time, a large number of consumer rights organizations in Europe and America are paying attention to this issue. Facebook and Google also ate tickets in Germany and France respectively. European regulators have embedded a set of principles in the “General Data Protection Regulations†that will come into effect in May 2018, which stipulates that biological information, including face lines, belongs to its owners. The use of this information requires the consent of the individual. The EU has achieved unprecedented heights in the protection and regulation of personal information.
"Human Face Turnstile" has been tested at the Capital Airport
Of course, some experts also pointed out that the current face recognition technology is still very rough, and the general public need not be too nervous. Overcorrecting will hinder the development of technology. In fact, in some cases, you can trick these systems into thinking that they see or hear something that doesn't actually exist.
The current Google researcher Alexey Kurakin and other artificial intelligence experts wrote that the neural network has security holes. As long as the image is changed very finely, the neural network will think that this image contains something that it doesn't actually have, and these changes are not felt by the naked human eye. Sometimes the change is only added anywhere in the picture. A few pixels. For example, an operator can change a few pixels of an elephant's photo and then trick the neural network into thinking it is "a car." Researchers call these "countermeasures."
However, even if the technical and regulatory aspects are still not mature, the broad application prospects of face recognition technology have been recognized by the business community.
The financial industry has sufficient identification requirements and has the broadest market prospects. It is estimated to be a "billion-dollar" market. In fact, China’s mobile payment companies are ahead of the world in terms of technology, security, and scenario requirements. They also have more technical requirements. At the same time, there are more than a dozen regulators related to identity certification. Everyone’s concerns and priorities are different. .
However, the Internet industry accustomed to "retreating regulatory reforms" has already begun to prepare its own standards, and has formed FIDO, which is generally accepted in the international market, Alipay-led IFAA, and the three authentication systems of WeChat Payment's SOTER. With the recent announcement of IFAA and SOTER’s ability to open up identity authentication, the three major authentication systems have started to play a positive game.
In social entertainment, face recognition technology is applied to the virtual character of the game, VR/AR social, expression interaction or short video, which inevitably produces a brand-new gameplay. For example, recently Facebook has a small game called FaceDance. This game is similar to a music game. The latter presses the correct key before the note reaches the recognition line, and FaceDance relies on face recognition, according to the game instructions. Out of all kinds of expressions. During the game, the camera will also record the entire process, allowing users to display their "Yan Yi" on social networking sites and upload various funny videos.
In terms of security, the Baidu AI robot has entered into cooperation with the Beijing Capital International Airport and has implemented it. The Baidu face gate has been placed at the Capital International Airport Transportation Control Center for testing, and is mainly responsible for the entrance and exit of office personnel and data monitoring. The International Civil Aviation Organization (ICAO) has determined that since 2010, its 118 member countries and regions must use machine-readable passports, and face recognition technology is the most important recognition model, which has become an international standard.
We can change mobile phones, and we can also fake identity cards and driver's licenses. With current medical technology, we cannot change our face. Face recognition technology has broad prospects and potential security risks. Will this technology be the beginning of a new human-computer interaction revolution or a fall in personal privacy? We will wait and see.
Titanium Round Bar is one of the most widely used titanium products. Currently Yesino production has covered the Gr. 1 Gr. 2 Gr. 3, Gr. 4, Gr. 5, Gr. 7 and Gr. 12, Gr. 23 general specification, etc. It meets different requirements of ASTM B348, ASTMF67/F136, ISO5832-2-3 and AMS 4928. Titanium Round applications include industrial, medical, aviation and other fields.Ti bars are also used in the moving parts of aircraft engines and propellers.
The most common applications for Titanium round bar, titanium hollow bar and titanium round rod include aerospace and aviation; petrochemical, gas & oil; the construction of water treatment plants; power generation facilities; and marine vehicles. Titanium`s light weight and extreme durability make it ideal for situations in which strength is required without adding to the overall mass of the finished product.
Our titanium bars find extensive applications across a wide range of industries. From aerospace and automotive to medical and industrial sectors, our products serve diverse needs. These bars are used in the manufacture of aircraft components, automotive parts, medical implants, marine equipment, and more. The versatility of our titanium bars opens up a world of possibilities for engineers, designers, and manufacturers, providing them with a high-performance material for their innovative solutions.
Titanium Round Bar ,Titanium Hollow Bar,Titanium Round Bar Stock,Titanium Round Rod
Yesino Metal Co., Ltd , https://www.yesinometal.com