Two dimensional < ahref=“ http://www.fuyingkeji.com/ "> main method of face recognition algorithm < / a >

Face recognition methods mainly focus on two-dimensional images. Two-dimensional face recognition mainly uses 80 nodes or punctuation points distributed on the face from low to high, and carries out identity authentication by measuring the distance between eyes, cheekbones and chin. Face recognition algorithms mainly include:

　　1. Method based on template matching: the template is divided into two-dimensional template and three-dimensional template. The core idea is to establish a three-dimensional adjustable model framework by using the law of human facial features. After locating the position of human face, the model framework is used to locate and adjust the position of human facial features, so as to solve the influence of observation angle, occlusion, expression change and other factors in the process of face recognition

　　2. Based on singular value feature method: the singular value feature of face image matrix reflects the essential attributes of the image, which can be used for classification and recognition

　　3. Subspace analysis: because of its strong description, low computational cost, easy implementation and good separability, it is widely used in face feature extraction and has become one of the mainstream methods of face recognition

　　4. Local preserving projection is a new subspace analysis method. It is the linear approximation of the nonlinear method laplacianeigenmap. It not only solves the disadvantage that traditional linear methods such as PCA are difficult to maintain the nonlinear manifold of the original data, but also solves the disadvantage that it is difficult for nonlinear methods to obtain the low-dimensional projection of new sample points

　　5. Principal component analysis (PCA)

PCA is an important method in the field of pattern recognition, which has been widely used in face recognition algorithms. Face recognition system based on PCA faces an important obstacle in application: incremental learning problem. It is very important for incremental PCA algorithm to reconstruct PCs from new samples, but with the increase of samples, this method needs to constantly abandon some unimportant PCs to keep the dimension of subspace unchanged, so the accuracy of this method is slightly poor

　　6. Other methods: elastic matching method, feature face method (based on KL transform), artificial neural network method, support vector machine method, integral image feature method (AdaBoost learning), probability model method