The system 1.5 seconds can recognize more than 3000 faces, even people who can move can recognize them.
This ugly face recognition system of Nanjing University of Science and Technology consists of an ordinary notebook computer and an equally ordinary camera. Although the "appearance" is not as high as expected, when the face recognition system really enters the "working state", its "working efficiency" is very high. After the employee's face image information is collected in advance, the employee only needs to walk to the camera of the system again, and the camera will immediately capture the employee's face information and transmit it to the system. According to the set requirements, the system will identify whether the facial features of the employee are consistent with the information collected in advance through database comparison. Only a few seconds later, the staff's face image appeared in the "hit record" column, and compared with the face image previously entered into the system, and finally confirmed the identity.
Li, director of the software engineering department of South University of Technology, revealed that the "working efficiency" of the "face recognition" system is calculated in seconds. The system can recognize 3,000-4,000 faces in 1.5 seconds, and can quickly recognize them among moving people.
Extracting facial features for "matching operation" is not afraid of "face-lifting"
Professor Li explained that the recognition function of the system is supported by the powerful "algorithm" of the computer. The "face recognition" system adopts special algorithms, which can realize face location, illumination processing and feature extraction, and finally achieve the purpose of comparison and recognition. The system seems to capture the face, but in fact, it is "staring" at some local information of the face, such as oval face or Chinese character face, how much the corner of the mouth rises, the geometric relationship of facial organs and so on. These are all captured by the system. "We will collect more than 80 feature points for comparison, so that the system can compare whether it is me or not according to the contour features of the face."
Can the face recognition system successfully match facial information after changing hair style, wearing glasses, or even more serious makeup removal and cosmetic surgery? Professor Li introduced that human factors such as basic hairstyle changes and makeup will not affect the working efficiency of the face recognition system. Because the system realizes the recognition function by comparing the local facial features of human faces. "If a person has no major facial injuries, such as disfigurement, in 15 years, his facial geometric features will not change much during this time interval. The system can recognize it. "
As for the problem that twins or plastic surgeons are more difficult to identify, Li said that a small operation like micro-plastic surgery only needs to input a large enough sample into the system for calculation, and then the identification will remain efficient. "We have done statistics, and the recognition rate of recent photos in normal experimental environment has exceeded 98%, while long photo matching, such as ID card matching, is quite different from recent photos, such as gaining weight, growing beard and changing hair style, which is close to 97% in normal experimental environment, reaching the leading level of the same kind in the world." Professor Li said.
■ Link
Some banks in Nanjing have begun to "brush their faces" to handle cards.
The "face recognition" system developed by Professor Li and his team has been put into use. This year, this "face recognition" system will enter the financial field. Some branches of a commercial bank in Jiangsu have started to use the system to handle debit card business by themselves, and they can do it by "brushing their faces" on the card issuing machine.
"Our next goal is to realize the large cash withdrawal of' brushing face' on ATM machines." Professor Li said that the team will strive to achieve this goal. After the householder is identified by the "manual identification" system, the ATM machine can give a "green light" to the large withdrawal of more than 50,000 yuan, thus realizing ATM self-service withdrawal. Greatly reduces the labor cost and saves time.
I hope I can help you, please adopt it, thank you!