Face detection

Face detection is one of the more widely used techniques of computer vision in practical applications, and it can be seen in many scenes. Face detection techniques combined with deep learning, especially convolutional neural networks, perform well in automatically extracting face features. Current research focuses on improving detection accuracy and speed, and enhancing model robustness. New advances in micro-expression recognition, model compression, anti-attack, face anti-fraud, etc. have driven the application of this technology in many fields, such as security monitoring, access control systems, and mobile payment.

With the continuous improvement of face recognition technology, more and more advanced tasks are beginning to use face recognition technology. For example, there are now multiple face recognition machines at the entrance of the train station, where travelers brush their ID documents while face matching is performed, and they can enter the station and take the train smoothly after passing the matching, which not only greatly saves manpower, but also reduces the possibility of stealing other people’s ID documents to enter the station and take the train.

The features of human face, like iris and fingerprints, have uniqueness, immutability, and non-replicability, thus laying the foundation for human identification.

Face image acquisition

The image needed for face recognition is the clear face image of human five senses, which can be obtained through video, motion pictures, pictures and other ways.

Pre-processing of face images

The collected images containing faces cannot be used directly for face recognition and need to be preprocessed. The image needs to be gray-scale transformed, filtered noise, sharpened and normalized and other processing.

Face feature extraction

The extraction of face features can be regarded as the key point localization of the image, and the position and size of the face can be judged by the position of the five senses of the person in the image. Face recognition is to pick out the useful feature information of the human face to realize face recognition, such as histogram features, structural features and so on.

Face feature extraction methods can be divided into two types: knowledge-based characterization methods and algebraic feature statistics-based characterization methods. The more commonly used is the knowledge-based characterization method, which calculates the relationship between the distance and angle of the eyes, nose, mouth and other features of the face through the positional structure of the face, and uses these structural relationships as the important features of face recognition.

Comparison and matching of face features

The face features to be recognized are searched and matched with the face features in the database, and when the similarity of the features reaches a set value, it is considered that the two have a large degree of similarity, thus realizing the face recognition task.

The commercialization of face recognition and other biometric technologies has brought a lot of convenience to people’s lives, but biological information belongs to the unchangeable and highly sensitive personal privacy, and there is a great security risk problem in the process of transmission, use and storage of these important data,. The over-analysis and misuse of these data without the person’s knowledge and consent is a serious violation of personal privacy, and the leakage of information will also lead to the disclosure of more important private information such as personal whereabouts.

Your task

What are the main steps in face detection?

Share your thoughts and ideas in the comments below.