1. The application field of face recognition
In fact, face recognition smart door technology has matured so far and is widely used in many fields. The security field is the earliest and most important application field of face recognition, which has penetrated into both civil security and public safety. In terms of civil security, face recognition is mainly used for access control and attendance systems in public places and home surveillance cameras. In terms of public safety, face recognition can be used for security checks in public places such as airports, and it is also an important intelligent case-handling method for real-time control of criminals and identity retrieval after the event. The financial field is also an important application field of face recognition, including remote bank account opening, smart pension collection, face-scanning payment, and a large number of identity authentication applications that ensure financial security.
In the field of transportation, face recognition is used as an important authentication method to ensure travel safety and improve traffic efficiency. At the same time, due to the convenience, high efficiency and contactless recognition of face recognition, there are also some Smart Door Lock brands that are equipped with face recognition technology on their products. Face recognition is also a type of biometric recognition, which uses a computer to analyze the face information of an image and identify its corresponding person identity. Face recognition technology is very similar to the human visual system. It obtains the feature points of the face from the image by understanding the characteristics of the face, so as to realize the identity confirmation. Users only need to make natural actions to complete collection and recognition, which is an indispensable ability for computers to realize intelligence.
Compared with other biometric identification, face recognition has the following advantages: (1) It is more convenient to collect. Compared with fingerprints, palm prints and irises, the physical scale of the human face is larger and the features are more obvious. (2) The feeling is more intuitive. People are very familiar with face images and can perceive a lot of information such as identity, but are generally unfamiliar with fingerprints, palm prints and iris images, and cannot directly determine identity and other information from these images. (3) The interaction is more insensitive. The face does not require the user to touch the device or actively cooperate during use, which is more hygienic and insensitive, while the use of fingerprints, palm prints and iris requires the user to touch the device or actively cooperate.
2. Principles of face recognition technology
While face recognition technology is widely used, ordinary people do not actually understand its internal logic and technical principles. This chapter will briefly introduce popular science to help you establish a correct understanding of face recognition.
The face recognition system is mainly composed of face detection, face key point location, face alignment, face feature extraction and recognition. Among them, face detection is the first step, which is responsible for locating the face from the image. area. If there is no face, no subsequent processing is performed.
Key point positioning is to locate key points such as the corners of the eyes, the tip of the nose, and the corners of the mouth on the detected face area to provide a basis for face alignment. Face alignment is based on the position information of key points, through affine transformation, the face area is normalized to a standard pose and a fixed size, and a standard face image is provided for face feature extraction.
Face feature extraction is the feature representation of the face image, the extraction of identity-related information contained in the face, and the suppression of identity-independent information, which is the basis of identity recognition.
Recognition is to judge or identify the identity of the face based on the features of the face. There are three kinds of recognition tasks: the first is judgment, that is, one-to-one authentication, to judge whether the face image is the person claimed, or to judge a pair of faces Whether the image is of the same person, the specific application is to provide ID card scenarios such as face ticket verification, remote account opening, etc. The second is selection, that is, one-to-many closed set recognition, input a face image, and select which person the face belongs to from the base library set. Specific applications include family member recognition and so on. The third is judgment + selection, that is, one-to-many open set recognition, input a face image, and judge whether the face belongs to the base library set. If so, select the person to which the face belongs. The specific applications include Blacklist or whitelist face recognition, face access control, etc. 3. The current technical difficulties of face recognition smart door locks Face recognition smart door locks carry out identification based on human facial feature information, collect feature values containing faces through cameras, automatically detect and track faces, and identify people to be identified. The features extracted from the face are discriminated according to the similarity. The main difficulty in the application of face recognition in smart door locks is that the application scenarios of door locks are relatively complex. The door lock is generally located in the aisle, and the light from different directions will make people's faces change greatly. When people approach the door lock quickly, the video faces will be distorted. This will lead to the deterioration of the quality of the collected face images, which will reduce the expression ability of the extracted face features.
The principle of 3D structured light unlocking is that the TrueDepth camera system composed of infrared lenses, floodlight sensing elements, etc., is projected on the user's face through more than 30,000 invisible light spots, creating a fine and complete image for the user's face. The 3D model is identified by emitting infrared photons in the IR module, and a unique "face spectrum" is drawn, while the flood light sensing element can help identify the user's face in the dark with the help of invisible infrared light. In addition, it is equipped with two independent high-performance chips, which can quickly respond to the read and recognized information. It can be confirmed by the algorithm to match the known "face spectrum", and the face recognition can be completed within 0.5s, and the unlocking efficiency is higher. . While being efficient and convenient, the 3D structured light algorithm has also obtained the BCTC certification. Because 3D structured light is three-dimensional, it can more accurately identify the user's biometrics and effectively identify hats, makeup, videos and other camouflages. The false recognition rate is only one in a million, far better than one in 50,000 fingerprints. To further ensure the security of the door lock.
