Facial recognition is a double-edged sword
Author: huifan Time: 2019-11-28
Facial treatment of “Chinese-style crossing” through face recognition, “brushing face” for identity verification first, face recognition toilet paper machine in public toilets ... unknowingly, the advanced face recognition technology gradually affects our life. Recently, US airport security screenings have to "brush their faces." So, what is the secret behind facial recognition?
Facial recognition in smartphones is an enhanced version of facial recognition technology. Using a computer for face recognition and identity verification can effectively simplify the company or government's access inspection process. Some systems can identify suspected people. The company can analyze the customer's facial characteristics to divide customer groups of different genders, ages and ethnic backgrounds, and formulate corresponding marketing strategies based on the division. Facial recognition technology can also serve consumers in specific fields, such as virtual glasses and virtual makeup.
Government agencies and companies can now easily trace each of us within the community, and even worldwide. Therefore, personal privacy issues are getting worse. The facial recognition market is currently worth about $ 3 billion and is expected to rise to $ 6 billion by 2021. The growing demand for surveillance is the main reason for its increased value, and the government is the main consumer of facial recognition systems.
About half of Americans have image information in the FBI database. In addition, network harassment based on facial recognition technology, and even the hidden dangers of reality tracking can not be ignored.
Today, facial recognition technology is becoming more common, and this emerging technology has gradually been used in the field of criminal investigations. Therefore, we should understand the "capable" and "impossible" of facial recognition technology, and understand how it recognizes our "face". Only then can we judge when and how to use facial recognition to recognize a person's "face".
How does "face recognition "work?
As part of a "biometric system", "facial recognition" distinguishes a person from others by monitoring its physical characteristics. Other types of work include the most common fingerprint matching, retinal scans, iris scans (scanning the more easily visible parts of the eye), and speech recognition.
These recognition systems first need to receive data from an "unknown person", usually image data; then the system analyzes the input data and attempts to match it with facial or sound characteristics of "known people" in the database. The recognition process of "face recognition" mainly has three steps: detection, "face pattern" creation, verification or recognition.
When you get a picture, the computer software analyzes it, and then identifies where the "face" appears in the picture, such as a crowd. For example, there are security cameras inside the mall, and computers equipped with facial recognition software will detect and identify video images in the mall.
Once the system finds any suspicious "faces", it will pay closer attention to each "face" in the picture. Recognition Sometimes it is necessary to adjust the orientation or size of the picture. Faces close to the camera may look slightly tilted or distorted; people farther away from the camera may appear small or even partially obstructed.
When the face recognition software adjusts the face in the image to a suitable size and orientation, it will perform more detailed recognition, and then create a so-called "face pattern". Much like a fingerprint record, a "face pattern" is actually a set of features that can distinguish a human face. A "face pattern" element includes the relative positions of facial features, such as the shape of eyes, eyebrows, and nose. A person with small eyes, thick eyebrows, and long and narrow noses is completely different from a person with large eyes, thin eyebrows, and wide noses. Eyes are the key to accurate facial recognition. A big black sunglasses can greatly confuse the facial recognition system, it can weaken the accuracy of facial recognition more than hairstyle or regular glasses.
You can verify the identity of a "known person" by comparing the "face pattern" with a photo, such as a company checking and verifying the identity of a person entering a secure area. Similarly, you can identify an "unknown person" by comparing the "face pattern" with a large number of pictures in the database.
"Face recognition" is not so easy
The key factor affecting the effect of facial recognition is light. A "face" that is evenly illuminated from the front, without shadows and any obscuration, is most easily identified. In addition, the intensity of the contrast between the face and the background of the image, and the distance of the face from the camera will greatly affect the effect of facial recognition.
Identifying people who consciously perform facial recognition will be simpler, such as employees who need to be authenticated to enter specific areas. Because they will cooperate with a higher degree of recognition, they will look directly at the camera in the appropriate light position, and adjust to the optimal position for easy software identification.
Others don't know they are recognized by a facial recognition system. In the image, it is more difficult to identify and analyze the faces of these people. The faces selected from the cluttered crowd must be digitally converted before generating "face patterns". Therefore, this is also a potential cause of errors in facial recognition software.
If a facial recognition system makes a mistake, there are a lot of potential problems. An access-restricted system may mistakenly allow access by an unauthorized person. On the contrary, the system may incorrectly restrict who is allowed to enter.
Surveillance cameras do not always capture clear pictures of suspects, which means that during police enforcement, facial recognition systems may recognize innocent people as suspects, or vice versa, to allow suspects to escape.
Generally, there are bridges in crime TV shows: police use facial recognition technology to quickly and accurately arrest suspects. But no matter how divine it is described in the play, the recognition technology will still go wrong in reality, although this recognition technology has been continuously improved. At present, the error rate of facial recognition is about 0.8%. According to the estimates of the National Institute of Science and Technology Standards, this error rate will drop by half every two years. The speech recognition error rate is more than 6%. In contrast, facial recognition is still more accurate. But the risk of facial recognition errors is still higher than that of iris scanning and fingerprint scanning.
The combined use of facial recognition technology and other technologies, including police camcorders, geolocation software, and machine learning to assist real-time tracking, goes beyond simple recognition. The combination of multiple technologies can determine a person's position, and also guess where they will go. Such a combination technology is very beneficial to combating crime, and it also makes our normal lives more privacy-free.
Technology provides us with powerful tools, and laws generally fail to keep pace with the development of new technologies. However, if we intend to continue to apply facial recognition technology to immigration policy and decision-making law enforcement, we must accept the benefits and losses brought by it, and we must understand the aspects of identification accuracy, personal privacy and ethics caused by this new technology The problem.
Linkedin: Huifan Technology
Youtube: Huifan Technology
Facebook: Huifan Technology