What is face recognition, what is the principle of face recognition technology, how it is realized step by step, what is the order of face recognition technology, and what is the recognition of face recognition. Today, I will help you popularize the principles and processes of face recognition technology.
The principle of face recognition technology is in fact a technology for biological discrimination of species. In a broad sense, it refers to the acquisition and detection of face images, data preprocessing, feature extraction, and matching recognition. A set of face recognition systems based on technical basics. With the passage of time, the face recognition technology has continuously developed and matured in the past two years. At present, more and more face recognition technology service providers are able to support the full range of face recognition technology services of the system.
The principle of face recognition technology is simple and convenient. The key points are three steps: one is that the unit must build a database containing a large number of face images; the other is to use various modes to win the target face image that needs to be identified at the moment; It compares and filters the target face image with the existing face image in the database. The technical process derived from the specific implementation of the principle of face recognition technology focuses on the following four parts, namely, the acquisition and monitoring of face images, the data preprocessing of face images, the collection of basic features of face images, and the face Image pairing and identification.
(1) Face image extraction and detection
The extraction and detection of the face image can be specifically divided into two locations: the extraction of the face image and the detection of the face image.
1. Acquisition of face images
There are two common channels for obtaining face images, namely, batch import of existing face images and extraction of face images at any time. Some of the more high-end face recognition systems can even be compatible and conditionally filter out face images that do not meet the requirements for face recognition quality or are of lower definition and quality, and extract them as clearly and accurately as possible.
Batch import of existing facial images: about to import the facial images obtained through various modes into the face recognition system in batches, and the system will automatically end the extraction business of face-by-face images.
Instant facial image extraction: the camera or high-definition camera is enabled to intelligently capture facial images at any time within the device's imageable range and terminate the extraction business.
2. Face image monitoring
A high-definition photo with a face image may contain other content under common conditions. At this time, it is necessary to carry out the necessary face detection. That is, in a face image, the system will accurately and accurately locate the position and size of the face, and at the same time as selecting useful image data, it will intelligently eliminate other redundant image data to effectively ensure the person. Accurate acquisition of face images.
(2) Data preprocessing of face images
The purpose of face image data preprocessing is to make deep processing of face images based on the basic conditions of the system's detection of face images to facilitate the feature extraction of face images.
The pre-processing of the face image specifically refers to the implementation of a series of cumbersome processing processes such as light, rotation, cutting, filtering, noise reduction, enlargement and reduction, etc., on the face image extracted by the system to make the face image no matter from the light. , Position, distance, size and other aspects can meet the standard requirements for feature extraction of face images.
(3) Feature extraction of face images
The features that the mainstream face recognition system can carry and use at present can be divided into the visual features of the face, the basic features of the pixel calculation of the face image, etc., and the feature extraction of the face image is collected for some specific basic features of the face. , The method of feature extraction is common and in view of the form of knowledge acquisition or the acquisition form based on algebraic characteristics.
Take the knowledge-based face recognition acquisition method as an example. Since the core of the face is composed of eyes, forehead, nose, ears, chin, mouth and other parts, it is possible to control this position including the mechanism relationship between them. The description is carried out with the basic characteristics of geometric shapes, that is to say, each person’s face image can have a corresponding geometric shape characteristic, which can help us to identify the basic characteristics of the major gaps in the face, which is also In view of the kinds of forms of knowledge acquisition.
(4) Automatic matching and discrimination of face images
We can set a face similarity level value in the face recognition system, and then compare the corresponding face image with all face images in the system database. If it exceeds the preset similarity value, the system will The excess face images will be output one by one. At this time, we need to implement fine screening based on the level of similarity of the face image and the original identity information data of the face. These precise screening processes can be divided into two categories: One is one-to-one screening, that is, to determine the sequence of face identities; the second is one-to-many screening, that is, the process of automatic matching and comparison in view of the similarity of faces.
The principle of face recognition technology is simple and convenient. It is to create a face image database, extract the target face image, and end the comparison and screening. In terms of the specific implementation sequence of the principle of face recognition technology, it is divided into four positions: acquisition and detection of face images, data preprocessing of face images, collection of basic features of face images, automatic matching and discrimination of face images. We can end the extraction business of facial images through two channels of batch importing existing facial images and extracting facial images at any time, and through a series of detection methods to make the facial image meet the specification requirements for feature extraction, and then After knowledge or other collection methods, we can obtain some important characteristics of facial image recognition. Later, we export facial images with facial similarity levels higher than certain conditions to facilitate our implementation of one-to-one and one-to-many accuracy. filter.