Thanks to the rapid development of mobile devices and digital cameras, face recognition technology advances, has become a major application support or configure a number of important products. In this paper, the current three technologies and its development direction and face recognition application a detailed introduction, hoping to provide a reference for all product development work.
Face recognition as a hot field of Internet technology, Internet products in many fields have a wide range of applications.
Many product managers often encounter the boss or department needs to work in one sentence: "We have to put forward a recognition of it." But recognition has been developing into a broader concept, providing different products through several different techniques application services. The different roles of people speak in different environments where face recognition, its products and technology behind expectations may be very different.
For example: Internet financial risk control recognition of the role you want, you may want to be on the face of the real-name authentication nuclear technology itself, or it may be based login authentication Face ID technology. The role of new media operators want to face recognition, may want to map changing face of marketing applications P based on face detection technology.
Understand the role and needs analysis, implementation is given basic skills product manager, if not familiar with the face recognition technology products and applications, it may not be the first time an accurate understanding of the needs of the role of the mind are thinking, delay project progress. The three will face recognition technology products and services provided to explain the analysis below.
First, Face Detection
Face detection means to identify a product's technology is part of the natural image area from a picture, also refer to the initial recognition significance. After years of technological improvements and research and development, has achieved recognition can detect multiple faces in a picture, as well as face detection in the video stream.
Most start face detection in the digital camera above, we all know that there is a portrait mode, Portrait mode is the basic principle of the image in the frame is processed on the previous digital cameras, recognizes an image area which belongs to the natural person and set the camera to focus on this area, so take out the faces in the picture is more clear. Face detection technology currently has two development directions as follows:
1. Face Recognition Properties
Face recognition refers to the property identified for the facial image area and then analyzed and processed to obtain images of the human face of a series of attributes such as gender, age, expression, race and so on. Applications mainly in rapidly build customer database portrait by portrait image and big data, enabling precision marketing.
2. facial features localization
Facial feature locating means for locating facial features is accurately identified the face image region, and to obtain the coordinate values of characteristic features of human face images. The product is the most mainstream of social commonly used products Images beauty and entertainment face transplant.
Second, the ratio of the face
Face refers to a measure of similarity than two faces of the picture, determine two faces in the picture are the same person natural kind of goods and services, technical services coupled with the need to use face detection to provide similar degree of similarity determination algorithm. Is also developed to the ratio of the video stream can be achieved and pictures comparison, as well as a variety of sensors face the incoming signal.
It should be noted that the face-matching technology can not guarantee 100% accuracy, it is necessary to set a threshold value on the product, above the threshold of similarity we judge people by face than (both determined to be the same person).
Face than the beginning of the application should be ancient arrest warrant, the warrant will usually attach a picture and characteristics of the offender, so that people together for face matching. In the criminal investigation techniques, by comparing the suspect's photograph and some evidence image, and helps to quickly identify the suspect's identity, promote the progress of the investigation of cases.
Currently using face contrast technology is the main direction of development:
1. Face retrieval
Face refers to the retrieval of the input face and the face in the database for comparison, a technique for determining the authentication input face belongs to a particular role, gates in the face, face recognition and other fields brush members product applications.
2. Face ID
Currently the most well-known face retrieval should be Apple's Face ID, Face ID realize 3D modeling and the head is stored in the phone through a variety of sensors, each landing will be unlocked through a variety of models to draw 3D image sensor and stored face alignment, in order to achieve logon and unlocking. According to publicity coincidence rate can reach the level of one part per million.
Third, in vivo detection
In the face matching technology into applications, will naturally encounter a problem: to compare the image or video stream face is real customer behavior, or falsely understood prosthesis (eg, using customer picture remake again) , in vivo detection technology has developed in such demand.
Vivo detection means determines whether the user operation is a real, against photos, videos, and other cheating mold attack. Currently technically divided into static and dynamic in vivo detection in vivo testing, the application of more dynamic in vivo testing.
Static vivo testing: refers to the judgment still pictures are real customer behavior or secondary remake, generally used in the attack defense is not high in the scene.
Dynamic vivo testing: refers to making behavior by instructing the user to verify whether the user is a real live my operation.
At present the main direction of development in vivo detection technology is the human face of nuclear body.
Face nuclear body
Vivo detection technologies described above face detection, face technology contrast, can achieve more reliable Internet authentication solutions.
Face Nuclear body refers to the user or for a selfie video from a camera, and citizenship information HD library passport, a 1: 1 face verification to confirm the identity of the user, and confirmed by the life detecting whether the current user himself and live.
Face nuclear body a wide range of product applications, once a year to verify the identity of the social security of the insured, the Internet real-name authentication system of lending, insurance sales of Internet real-name authentication, banking / securities accounts, and so remote.
Conclusion: the technical and product applications described above, the current BAT companies have an open platform to provide technical interface support, you can choose their own partner product technologies for product design can also contact me to communicate (non smart thinking pill)