Knowledge background:
With the continuous progress of society and the urgent requirements for fast and effective automatic identity verification in all aspects, biometric identification technology has developed rapidly in recent decades. As an inherent attribute of human beings with strong self-stability and individual differences, biometrics have become the most ideal basis for automatic identity verification. The current biometric identification technologies mainly include: fingerprint identification, retinal identification, iris identification, gait identification, vein identification, face identification, etc. Compared with other recognition methods, face recognition has been widely researched and applied because of its direct, friendly and convenient features, users have no psychological barriers, and it is easy to be accepted by users. In addition, we can further analyze the results of face recognition, and obtain many additional rich information about people's gender, expression, age, etc., which expands the application prospects of face recognition.1 Face Detection
Given an image, find out the positions of all faces in the image, usually framed by a rectangular frame, the input is an image img, and the output is several rectangular frame positions (x, y, w, h) containing faces, like this.
From the domain of the problem, face detection belongs to the field of object detection.
Automatic face detection is the foundation of all applications around automatic face image analysis, including but not limited to: face recognition and verification, face tracking in surveillance situations, facial expression analysis, facial attribute recognition (gender/age recognition, face value evaluation), facial lighting adjustment and deformation, facial shape reconstruction, image video retrieval, organization and presentation of digital photo albums. Face detection is the initial step in all modern vision-based human-computer, and human-robot, interaction systems.
2 Face Identification (Face Identification)
Face recognition is the process of classifying known faces. Face recognition usually includes the following four steps:
1) Face detection: is the process of locating face regions in an image. (That is, caring about whether it is a human face or not, and caring about who is not the task of this step)
2) Face preprocessing: Adjust the face image to make it look clearer.
3) Collect and learn faces: collect many pre-processed faces (these faces are the faces of the people to be recognized, not others) , and then learn how to recognize them.
4) Face Recognition: Find out which face is most similar to the face in the camera among the collected faces.
Identification principle:
The existing face recognition methods mainly include deep learning and non-deep learning methods, which essentially extract the features of face images, and combine the extracted features with existing feature templates or models (equivalent to training a classifier) to compare and find the one with the greatest similarity.
1 Feature extraction
We hope that when extracting features of a face, for the same sample (face), its features can remain relatively stable under various lighting conditions or angles. Therefore, there are many special extraction methods: feature points, PCA dimensionality reduction, LBP features, etc.
LBP
2 Classifier training
Deep learning: Classification uses neural networks for classification, various neural networks. (universal)
Non-deep learning: classification methods such as SVM, template matching, etc. (with particularity)
3 Face Recognition
Input the classifier and output the recognition result.
Code
Minimum distance method (matlab)
NearestNeighbor（python&matlab））
LBP-DBN (matlab)
FaceRecognition-tensorflow python）
Summarize:
Sort out the principles of face recognition and the functions of each part, and introduce several basic methods of face recognition and code implementation.